(Arno Bohm (Auth.) ) Quantum Mechanics Foundations (B-Ok - Xyz)

Texts and
Monographs
in Physics
w. Beiglbock
J. L. Birman
E. H. Lieb
T. Regge
W. Thirring
Series Editors
Texts and Monographs in Physics
R. Bass: Nuclear Reactions with Heavy Ions (1980).

A. Bohm: Quantum Mechanics: Foundations and Applications, Second Edition
(1986).
O. Bratteli and D. W. Robinson: Operator Algebras and Quantum Statistical
Mechanics. Volume I: C*- and W*-Algebras. Symmetry Groups. Decomposition
of States (1979). Volume II: Equilibrium States. Models in Quantum Statistical
Mechanics (1981).
K. Chadan and P.C. Sabatier: Inverse Problems in Quantum Scattering Theory
(1977).
M. Chaichian and N.F. Nelipa: Introduction to Gauge Field Theories (1984).
G. Gallavotti: The Elements of Mechanics (1983).
W. Glockle: The Quantum Mechanical Few-Body Problem (1983).
W. Greiner, B. Muller, and J. Rafelski: Quantum Electrodynamics of Strong Fields
(1985).
J.M. Jauch and F. Rohrlich: The Theory of Photons and Electrons: The Relativistic
Quantum Field Theory of Charged Particles with Spin One-half, Second
Expanded Edition (1980).
J. Kessler: Polarized Electrons (1976). Out of print. (Second Edition available as
Springer Series in Atoms and Plasmas, Vol. I.)
G. Ludwig: Foundations of Quantum Mechanics I (1983).
G. Ludwig: Foundations of Quantum Mechanics II (1985).
R.G. Newton: Scattering Theory of Waves and Particles, Second Edition (1982).
A. Perelomov: Generalized Coherent States and Their Applications (1986).
H. Pilkuhn: Relativistic Particle Physics (1979).
R.D. Richtmyer: Principles of Advanced Mathematical Physics. Volume I (1978).
Volume II (1981).
W. Rindler: Essential Relativity: Special, General, and Cosmological, Revised
Second Edition (1980).
P. Ring and P. Schuck: The Nuclear Many-Body Problem (1980).
R.M. Santilli: Foundations of Theoretical Mechanics. Volume I: The Inverse
Problem in Newtonian Mechanics (1978). Volume II: Birkhoffian Generalization
of Hamiltonian Mechanics (1983).
M.D. Scadron: Advanced Quantum Theory and Its Applications Through
Feynman Diagrams (1979).
N. Straumann: General Rl!lativity and Relativistic Astrophysics (1984).
C. Truesdell and S. Bharatha: The Concepts and Logic of Classical
Thermodynamics as a Theory of Heat Engines: Rigourously Constructed
upon the Foundation Laid by S. Carnot and F. Reech (1977).
F.J. Ynduniin: Quantum Chromodynamics: An Introduction to the Theory of
Quarks and Gluons (1983).
Arno Bohm
QuantUITl Mechanics:
F oundations and
Applications
Second Edition, Revised and Enlarged

Prepared with M. Loewe
With 94 Illustrations
6 Springer Science+Business Media, LLC

Arno Bohm
Department of Physics
Center for Particle Theory
The University of Texas at Austin
Austin, TX 78712
U.S.A.
Editors
Wolf Beiglbock Joseph L. Birman
Institut fUr Angewandte Mathematik Department of Physics
Universităt Heidelberg The City College of the
Im Neuenheimer Feld 5 City University of New York
D-6900 Heidelberg 1 New York, NY 10031
Federal Republic of Germany U.S.A.
Elliott H. Lieb Tullio Regge Walter Thirring

Department of Physics Istituto de Fisica Teorica Institut fUr Theoretische PIiysik
losepIi Henry Laboratories Universita di Torino der Universităt Wien
Princeton University C. so M. d'Azeglio, 46 Boltzmanngasse 5
Princeton, Nl 08540 10125 Torino A-1090 Wien
U.S:A. Italy Austria
Library of Congress Cataloging in Publieation Data

Bohm, Amo, 1936~
Quantum meehanies.
(Texts and monographs in physies)
Bibliography: p.
Includes index.
1. Quantum theory. 1. Title. II. Series
QCI74.12.B63 1986 530.1'2 85-4710
The first edition ofthis book appeared as: Amo Bohm, Quantum Mechanics. Springer-
Veriag, New York, Heidelberg, Berlin, 1979.
© 1979, 1986 by Springer Science+Business Media New York

Originally published by Springer-Verlag New York Ine. in 1986
Softcover reprint ofthe hardcover 2nd edition 1986
AlI rights reserved. No part of this book may be translated or reprodueed in any
form without written permission from Springer-Verlag, 175 Fifth Avenue, New York,
New York 10010, V.S.A.
Typeset by Composition House Ltd., Salisbury, England.
9 8 7 6 5 4 3 2
ISBN 978-3-540-13985-0 ISBN 978-3-662-01168-3 (eBook)

DOI 10.1007/978-3-662-01168-3
To my students, colleagues, and friends
without whose help this second edition
would not have been possible.
Preface
The first edition of this book was written as a text and has been used many
times in a one-year graduate quantum mechanics course. One of the
reviewers has made me aware that the book can also serve as, " ... in
principle, a handbook of nonrelativistic quantum mechanics." In the second
edition we have therefore added material to enhance its usefulness as a
handbook. But it can still be used as a text if certain chapters and sections are
ignored. We have also revised the original presentation, in many places at the
suggestion of students or colleagues. As a consequence, the contents of the
book now exceed the material that can be covered in a one-year quantum
mechanics course on the graduate level. But one can easily select the material
for a one-year course omitting-according to one's preference-one or
several of the following sets of sections: {1.7, XXI}, {X, XI} or just {XI}, {II.7,
XIII}, {XIV.5, XV}, {XIX, XX}. Also the material of Sections 1.5-1.8 is not
needed to start with the physics in Chapter II. Chapters XI, XIII, XIX, and
XX are probably the easiest to dispense with and I was contemplating the
deletion of some of them, but each chapter found enthusiastic supporters
among the readers who advised against it.
Chapter I-augmented with some applications from later chapters-can
also be used as a separate introductory text on the mathematics of quantum
mechanics.
The book is self-contained and does not require any prior knowledge of
quantum mechanics, but it is a difficult book, because it is so concise. It offers
a huge amount of material, more than one can find in texts with twice the
number of pages. Consequently, some familiarity with the subject would be
vii
viii Preface
very helpful. Prerequisites are a knowledge of calculus, vector algebra, and

analysis. Most physical examples are taken from the fields of atomic and
molecular physics, as it is these fields that are best known to students at the
stage when they learn quantum mechanics.
Texts on a subject established a half-century before are often written using
the material and presentation established by the first generation of books on
that subject. New applications, deeper insights, and unifying formulations
that subsequently develop are easily overlooked. That has not been done in
this text. I have presented a unified theoretical formulation which was made
possible by later developments, and have included examples from more
recent papers.
The changes incorporated in the second edition provide an easier access to
the material, but leave the general idea unchanged. It is therefore fitting to
quote from the preface of the first edition:
... in contrast to what one finds in the standard books, quantum

mechanics is more than the overemphasized wave-particle dualism
presented in the familiar mathematics of differential equations. "This
latter dualism is only part of a more general pluralism" (Wigner)
because, besides momentum and position, there is a plurality of
other observables not commuting with position and momentum. As
there is no principle that brings into prominence the position and
momentum operators, a general formalism of quantum mechanics,
in which every observable receives the emphasis it deserves for the
particular problem being considered, is not only preferable but often
much more practical .... It is this general form of quantum theory
that is presented here.
I have attempted to present the whole range from the fundamental
assumptions to the experimental numbers. To do this in the limited
space available required compromises. My choice ... was mainly
influenced by what I thought was needed for modern physics and by
what I found, or did not find, in the standard textbooks. Detailed
discussions of the Schrodinger differential equation for the hydrogen
atom and other potentials can be found in many good books.! On
the other hand, the descriptions of the vibrational and rotational
spectra of molecules are hardly treated in any textbooks of quantum
mechanics, though they serve as simple examples for the important
procedure of quantum-mechanical model building .... So I have
treated the former rather briefly and devoted considerable space to
the latter.
Groups have not been explicitly made use of in this book.
However, the reader familiar with this subject will see that group
theory is behind most of the statements that have been cast here in
terms of algebras of observables.
1 The subject also is usually adequately treated in undergraduate courses.

Preface ix
This is a physics book, and though mathematics has been used

extensively, I have not endeavored to make the presentation math-
ematically rigorous .... Except in the mathematical inserts, which
are given in openface brackets [M: ], the reader will not even be
made aware of these mathematical details.
The mathematical inserts are of two kinds. The first kind provides
the mathematics needed, and the second kind indicates the under-
lying mathematical justification ....
Quantum mechanics starts with Chapter II, where the most
essential basic assumptions (axioms) of quantum mechanics are
made plausible from the example of the harmonic oscillator as
realized by the diatomic molecule. Further basic assumptions are
introduced in later chapters when the scope of the theory is extended.
These basic assumptions ("postulates") are not to be understood as
mathematical axioms from which everything can be derived without
using further judgment and creativity. An axiomatic approach of this
kind does not appear to be possible in physics. The basic assump-
tions are to be considered as a concise way of formulating the
quintessence of many experimental facts.
The book consists of two clearly distinct parts, Chapters II-XI
and Chapters XIV-XXI, with two intermediate chapters, Chapters
XII and XIII. The first part is more elementary in presentation,
though more fundamental in subject matter .... The second part,
which starts with Chapter XIV, treats scattering and decaying
systems. The presentation there is more advanced.
Chapter XIV gives a derivation of the cross section under very
general conditions .... Two different points of view-one in which
the Hamiltonian time development is assumed to exist, and the other
making use of the S-matrix-are treated in a parallel fashion. The
required analyticity of the S-matrix is deduced from causality. One of
the main features of the presentation is to treat discrete and
continuous spectra from the same point of view. For this the rigged
Hilbert space is needed, which provides not only a mathematical
simplification but also a description which is closer to physics.
Major changes to the book have been made in Chapters I, XIII, and XXI,
which were almost totally rewritten. Chapter XXI discusses the new notion of
Gamow vectors for the description of decaying states. They were created
when the first edition was written in order to achieve the desired unity of
description of all of quantum mechanics. Chapter I had to be expanded to
provide the mathematical background for Chapter XXI. To start a physics
book with a mathematical introduction may create an incorrect impression. I
therefore want to emphasize that the book contains many more experimental
numbers than mathematical theorems. Extensive revisions have also been
made in Chapters II, IV, XIV, XVI, XVII, and XVIII; and many improve-
ments were made in Chapters III, V, VIII, and IX. The appendix to Section
x Preface
V.3 has been rewritten to provide a simple but typical example for the
construction of noncom pact group representations. Not all chapters could be
revised because of time limitations. Chapters VII, X, XI, and XIX have been
scrutinized only a little and Chapters VI, XII, XV, and XX remain essentially
as they were in the first edition.
Acknowledgments
For the second edition, as for the first edition, I am indebted to many for their
help, encouragement, and advice. Chapter XIII was rewritten with K. Kraus,
who together with A. Peres also suggested improvements to Chapter II. On
the material of Chapter XXI, I received advice from L. Khalfin and M.
Gadella. The new version of Chapter I grew out of a joint project with G. B.
Mainland. The revisions of the first part of the book were made together with
M. Loewe. For the revisions of the second part of the book I was assisted by J.
Morse. P. Busch proofread Section XIII.l. I received many letters pointing
out misprints and inadequacies, suggesting improvements, and encouraging
me through the tedious task of preparing a new edition. I would like to thank
R. Scalettar, A. Y. Klimik, L. Fonda, and T. Mertelmeier for pointing out
errors in the first edition. The numerous misprints could not have been
purged without the help of students in my classes. Support from D.O.E. and
the Alexander von Humboldt Foundation is gratefully acknowledged. I am
particularly grateful to M. Loewe who proofread the entire book and made
many improvements. If the second edition is better than the first, it is mainly
due to him.
xi
Contents
CHAPTER I
Mathematical Preliminaries
1.1 The Mathematical Language of Quantum Mechanics 1
1.2 Linear Spaces, Scalar Product 2
1.3 Linear Operators 5
1.4 Basis Systems and Eigenvector Decomposition 8
1.5 Realizations of Operators and of Linear Spaces 18
1.6 Hermite Polynomials as an Example of Orthonormal Basis Functions 28
Appendix to Section 1.6 31
1.7 Continuous Functionals 33
1.8 How the Mathematical Quantities Will Be Used 39
Problems 39
CHAPTER II
Foundations of Quantum Mechanics-The Harmonic Oscillator 43
II. I Introduction 43
11.2 The First Postulate of Quantum Mechanics 44
11.3 Algebra of the Harmonic Oscillator 50
11.4 The Relation Between Experimental Data and Quantum-Mechanical
Observables 54
11.5 The Basic Assumptions Applied to the Harmonic Oscillator, and
Some Historical Remarks 74
11.6 Some General Consequences of the Basic Assumptions of Quantum
Mechanics 81
11.7 Eigenvectors of Position and Momentum Operators; the Wave
Functions of the Harmonic Oscillator 84
X1ll
XIV Contents
II.8 Postulates II and III for Observables with Continuous Spectra 94

II.9 Position and Momentum Measurements-Particles and Waves 101
Problems 112
CHAPTER III
Energy Spectra of Some Molecules 117
III.1 Transitions Between Energy Levels of Vibrating Molecules-
The Limitations of the Oscillator Model 117
III.2 The Rigid Rotator 128
III.3 The Algebra of Angular Momentum 132
III.4 Rotation Spectra 138
III.5 Combination of Quantum Physical Systems-The Vibrating Rotator 146
Problems 155
CHAPTER IV
Complete Systems of Commuting Observables 159
CHAPTER V
Addition of Angular Momenta-The Wigner-Eckart Theorem 164
V .1 Introduction-The Elementary Rotator !64
V.2 Combination of Elementary Rotators 165
V.3 Tensor Operators and the Wigner-Eckart Theorem 176
Appendix to Section V.3 181
V.4 Parity 192
Problem 204
CHAPTER VI
Hydrogen Atom-The Quantum-Mechanical Kepler Problem 205
VI.I Introduction 205
VI.2 Classical Kepler Problem 206
VI.3 Quantum-Mechanical Kepler Problem 208
VI.4 Properties of the Algebra of Angular Momentum and the Lenz Vector 213
VI.5 The Hydrogen Spectrum 215
Problem 222
CHAPTER VII
Alkali Atoms and the Schr6dinger Equation of
One-Electron Atoms 223
VII.! The Alkali Hamiltonian and Perturbation Theory 223
VII.2 Calculation of the Matrix Elements of the Operator Q -, 227
VII.3 Wave Functions and Schrodinger Equation of the Hydrogen Atom
and the Alkali Atoms 234
Problem 241
CHAPTER VIII
Perturbation Theory 242
VIlLI Perturbation of the Discrete Spectrum 242
VIII.2 Perturbation of the Continuous Spectrum-
The Lippman-Schwinger Equation 248
Problems 251
Contents xv
CHAPTER IX
Electron Spin 253
IX.l Introduction 253
IX.2 The Fine Structure-Qualitative Considerations 255
IX.3 Fine-Structure Interaction 261
IX.4 Fine Structure of Atomic Spectra 268
IX.5 Selection Rules 270
IX.6 Remarks on the State of an Electron in Atoms 271
Problems 272
CHAPTER X
Indistinguishable Particles 274
X.l Introduction 274
Problem 281
CHAPTER XI
Two-Electron Systems- The Helium Atom 282
XL! The Two Antisymmetric Subspaces of the Helium Atom 282
XL2 Discrete Energy Levels of Helium 287
XI.3 Selection Rules and Singlet-Triplet Mixing for the Helium Atom 297
XL4 Doubly Excited States of Helium 303
Problems 309
CHAPTER XII
Time Evolution 310
XII.l Time Evolution 310
XII.A Mathematical Appendix: Definitions and Properties of Operators
that Depend upon a Parameter 324
Problems 326
CHAPTER XIII
Some Fundamental Properties of Quantum Mechanics 328
XIII. I Change of the State by the Dynamical Law and by the
Measuring Process-The Stern-Gerlach Experiment 328
Appendix to Section XIII. 1 340
XIII.2 Spin Correlations in a Singlet State 342
XIII.3 Bell's Inequalities, Hidden Variables, and the Einstein-Podolsky-
Rosen Paradox 347
Problems 354
CHAPTER XIV
Transitions in Quantum Physical Systems-Cross Section 356
XIV.! Introduction 356
XIV.2 Transition Probabilities and Transition Rates 358
XIV.3 Cross Sections 362
XIV.4 The Relation of Cross Sections to the Fundamental Physical
Observables 365
XIV.5 Derivation of Cross-Section Formulas for the Scattering of
a Beam off a Fixed Target 368
Problems 384
xvi Contents
CHAPTER XV
Formal Scattering Theory and Other Theoretical Considerations 387
XV. I The Lippman-Schwinger Equation 387
XV.2 In-States and Out-States 391
XV.3 The S-Operator and the Melller Wave Operators 399
XV.A Appendix 407
CHAPTER XVI
Elastic and Inelastic Scattering for Spherically Symmetric
Interactions 409
XVI.I Partial-Wave Expansion 409
XVI.2 Unitarity and Phase Shifts 417
XVI.3 Argand Diagrams 422
Problems 424
CHAPTER XVII
Free and Exact Radial Wave Functions 425
XVII.I Introduction 425
XVII.2 The Radial Wave Equation 426
XVII.3 The Free Radial Wave Function 430
XVII.4 The Exact Radial Wave Function 432
XVII.5 Poles and Bound States 439
XVII.6 Survey of Some General Properties of Scattering Amplitudes and
Phase Shifts 441
XVII. A Mathematical Appendix on Analytic Functions 444
Problems 450
CHAPTER XVlIl
Resonance Phenomena 452
XVIII. I Introduction 452
XVIII.2 Time Delay and Phase Shifts 457
XVIII. 3 Causality Conditions 464
XVIII.4 Causality and Analyticity 467
XVIII.5 Brief Description of the Analyticity Properties of the S-Matrix 471
XVIII.6 Resonance Scattering-Breit-Wigner Formula for Elastic Scattering 476
XVIII. 7 The Physical Effect of a Virtual State 487
XVIII. 8 Argand Diagrams for Elastic Resonances and Phase-Shift Analysis 489
XVIII.9 Comparison with the Observed Cross Section: The Effect of
Background and Finite Energy Resolution 493
Problems 503
CHAPTER XIX
Time Reversal 505
XIX.l Space-Inversion Invariance and the Properties of the S-Matrix 505
XIX.2 Time Reversal 507
Appendix to Section XIX.2 511
XIX.3 Time-Reversal Invariance and the Properties of the S-Matrix 512
Problems 516
Contents XVll
CHAPTER XX
Resonances in Multichannel Systems 517
XX.! Introduction 517
XX.2 Single and Double Resonances 518
XX.3 Argand Diagrams for Inelastic Resonances 532
CHAPTER XXI
The Decay of Unstable Physical Systems 537
XXI.I Introduction 537
XXI.2 Lifetime and Decay Rate 539
XXI.3 The Description of a Decaying State and the Exponential Decay Law 542
XXI.4 Gamow Vectors and Their Association to the Resonance Poles of the
S-Matrix 549
XXl.5 The Golden Rule 563
XXI.6 Partial Decay Rates 567
Problems 569
Epilogue 571
Bibliography 574
Index 579
CHAPTER I
Mathematical Preliminaries
The mathematical language of quantum mechanics is introduced in this

chapter. It does not contain any physics.
1.1 The Mathematical Language of Quantum Mechanics

To formulate Newtonian mechanics, the mathematical language of differ-
ential and integral calculus was developed. Though one can get some kind
of understanding of velocity, acceleration, etc., without differential calculus
(in particular for special cases), the real meanings of these physical notions
in their full generality become clear only after one is familiar with the idea
of the derivative. On the other hand, though, the abstract mathematical
definitions of calculus become familiar to us only if we visualize them in
terms of their physical realizations. Nowadays, no one would attempt to
understand classical mechanics without knowing calculus.
Quantum mechanics, too, has its mathematical language, whose develop-
ment went parallel to the development of quantum mechanics and whose
creation in its full generality was inspired by the needs of quantum physics.
This is the language of linear spaces, linear operators, associative algebras,
etc., which has meanwhile grown into one of the main branches of mathe-
matics-linear algebra and functional analysis. Although one might obtain
some sort of understanding of quantum physics without knowing its mathe-
matical language, the precise and deep meaning of the physical notions in
their full generality will not reveal themselves to anyone who does not under-
stand its mathematical language.
2 I Mathematical Preliminaries
Therefore we shall start the quantum-mechanics course with some of the

vocabulary and grammar of this language. We shall not try to be mathe-
matically rigorous, since one can still communicate in a language that one
does not speak completely correctly. We shall also not give all the mathe-
matics that is needed at the beginning, and you need not be worried if you
do not understand everything right away; one learns a language best by
using it. We shall give in this chapter not much more mathematics than is
needed to make the initial statements about physics. We shall then have to
learn new mathematical notions whenever they arise, while we proceed with
the development of the physical ideas.
Before we start to study the mathematical structures that are employed
in quantum mechanics, we should make the following observation: A
mathematical structure is not something real-it only exists in our mind
and is created by our mind (though often inspired by outside influences). It
is obtained by taking a set of objects and equipping this set with a structure
by defining relations between these objects. Modern mathematics distin-
guishes three basic kinds of structures: algebraic, topological, and ordering.
The mathematical structures we use are complicated combinations of these
three. For example, the real numbers have an algebraic structure given by
the usual laws of addition and multiplication; they have a topological
structure given by the meaning of the usual limiting process for an infinite
series of numbers, and they have an ordering structure given by the relations
expressed by <.
We shall use predominantly algebraic structures, although in order to
speak the mathematical language of quantum mechanics correctly, topo-
logical structures are essential. We shall start with the definition of a linear
space, and linear operators and give the definition of an associative algebra.
That will provide us with enough vocabulary and grammar to enable us to
start communicating physics. We will then give an intuitive description of
some fundamental mathematical properties needed throughout the book.
In Section 1.4 we present the eigenvector decomposition. The underlying
mathematics is the nuclear spectral theorem, one of the greatest achievements
of mathematics. We shall not prove it, we shall not even precisely state it, but
explain it by analogy as a generalization of the coordinate expansion in the
three-dimensional case. Section 1.5 describes the spaces of sequences and
functions as the coordinate representation of linear spaces (in which the
nuclear spectral theorem holds). The generalized eigenvectors, of which the
reader should have acquired an intuitive understanding in Sections 1.4 and
1.5, will then be defined as continuous functionals in Section 1.7, which is the
only place in which a brief discussion of some topological notions is given.
1.2 Linear Spaces, Scalar Product
A linear space is a set of elements c/J, 1/1, x, ... which is given an algebraic
structure that is a generalization of certain aspects of the three-dimensional
real space 1Il 3 . The elements, also called vectors, are defined to obey rules
1.2 Linear Spaces, Scalar Product 3
which are well-known properties of vectors in [R3. The linear spaces which
we define are in general not three-dimensional, but can have any dimension
N, often infinite, and use, in general, the complex numbers C rather than the
real numbers [R. The rules that define the linear space are:
(a) For any two elements cj>, 1/1 E «I> there is defined an element cj> + 1/1 E «1>,
the sum or addition of cj> and 1/1, which obeys the following rules:
cj>+ 1/1 = 1/1 + cj>, (2.1 a)
(cj> + 1/1) + X = cj> + (1/1 + X)· (2.1b)
There exists in «I> an element O-called the zero
vector-with the property cj> + 0 = cj> for every cj> E «1>. (2.1c)
(b) For every complex number, a E C, and every element cj> E «1>, there
exists an element acj> E «1>, called the product of the vector cj> with the
number a, which obeys the following rules:
lcj> = cj>, (2.1d)
Ocj> = O. (2. Ie)
(0 on the left is the number zero, 0 on the right is the zero vector of
(2.1c).)
a(bcj» = (ab)cj>, a,b E C, (2.1f)
a(cj> + 1/1) = acj> + al/l, (2.1g)
(a + b)cj> = acj> + bcj>. (2.1 h)
The element (-I)cj> is usually denoted -cj>; because of (2.1d), (2.1h),
and (2. Ie) one obtains the following result:
cj> + (-cj» = (1 + (-l)cj> = Ocj> = O.
The above collection of relations between elements of «I> and complex
numbers defines the linear space and every set of objects that fulfills these
relations is called a linear space. Often such objects have more properties
than those stated above.
Clearly the vectors a, b, ... in the three-dimensional space [R3 fulfill the
above relations. However, the set of complex infinitely differentiable con-
tinuous functions which vanish rapidly at infinity also fulfills these relations.
One often says that the abstract linear space defined by the above rules is
realized by other mathematical objects like functions, if these other objects
appear to us more" real" than the" abstract" vectors. Thus if one feels more
familiar with functions one may prefer the "realization" of «I> by a space of
functions over the space «I> itself.
In physics the abstract mathematical objects are realized by physical
objects. Thus a physicist's "realization" of a linear space is not by other
more familiar or more interesting mathematical objects, but by physical
objects. In particular, in quantum physics, the elements of the space «I> will
be the mathematical images of pure physical states; thus, the linear space is
"realized" by the physical states of a quantum system.
A linear space is a set with very little mathematical structure. We will

equip it with more structure by defining a scalar product. This notion is
again a generalization from the three-dimensional real space 1R3.
A linear space is called a scalar product space (or Euclidean space or
pre-Hilbert space) ifin it a function (cp, ljJ) of the two vectors cp, ljJ E is defined
which is a complex number and has the following properties
(cp, cp) 2:: 0 and (cp, cp) = 0 iff cp = o. (2.2a)
~~=~~ ~~
(the bar denotes complex conjugate).
(cp, aljJ) = a(cp, ljJ) (2.2c)
(a E C, the set of complex numbers).
(CPl + CP2' ljJ) = (cpp ljJ) + (CP2' ljJ). (2.2d)
This function is called the scalar product of the elements cp and ljJ. The
usual scalar product in 1R 3 , (a, b) = a . b clearly fulfills the conditions
(2.2a)-(2.2d) with all numbers being real instead of complex.
As in 1R3 one calls two vectors cp and ljJ orthogonal if
(cp, ljJ) = o. (2.3)
With the scalar product defined by (2.2a)-(2.2d) one defines the norm
I cp Iof a vector cp by
Ilcpll = +J(CP, cp). (2.4)
For any vector ljJ different from the zero vector one can always define a
vector ~ = ljJ/llljJll, which has the property II~II = 1 and is called a normalized
vector.
Sometimes one needs in a linear space a more general notion than the
scalar product, the bilinear Hermitian form.
A complex-valued function h(cp, ljJ) of two vector arguments is a Hermitian
form if it satisfies
h(cp, ljJ) = h(ljJ, cp), (2.5b)
h(cp, aljJ) = ah(cp, ljJ) (a E C), (2.5c)
h(CPl + CP2' ljJ) = h(CP1' ljJ) + h(CP2' ljJ). (2.5d)
If in addition h satisfies
h(cp, cp) 2:: 0 (2.5a)
for every vector cp, then h is said to be a positive Hermitian form. A positive
Hermitian form is called positive definite if
from h( cp, cp) = 0 follows cp = 0 for every vector cp. (2.6)
Thus a Hermitian form fulfills the conditions (2.2b), (2.2c), and (2.2d), but
not the condition (2.2a) for a scalar product. However, a positive definite
Hermitian form is, by (2.6), a scalar product.
Positive Hermitian forms, which are not necessarily scalar products,

satisfy the Cauchy-Schwarz-Bunyakovski inequality:
Ih(<p, 1/1)1 2 :::;; h(<p, <p)h(l/I, 1/1). (2.7)
If h is positive definite, equality holds iff is called a subspace of if M is a linear space
under the same definitions of the operations of addition and multiplication
by a number as given for <1>, i.e., if it follows from <p, 1/1 E M that a<p E M and
<p + 1/1 E M.
An expression of the form a 1 <p 1 + an <Pn is called a linear combination of
the vectors <Pl' <P2"'" <Pn; the vectors <Pl' <P2"'" <Pn are said to be linearly
dependent if there exist numbers a 1 , a2, ... , an' not all zero, for which
a 1 <Pl + a2<P2 + ... + an<pn = O. If the equation a 1 <Pl + a 2<P2 + ...
+ an<pn = 0 holds only for a1 = a2 = ... = an = 0, then the vectors <Pl'
<P2' ... , <Pn are called linearly independent. A space is said to be finite-
dimensional and, more precisely, n-dimensional if there are n and not more
than n linearly independent vectors in <1>. If the number of linearly indepen-
dent vectors in is arbitrarily great, then is said to be infinite-dimensional.
Every system of n linearly independent vectors in an n-dimensional space 
is called a basis for <1>.
An isomorphism between two algebraic structures d and f!J is a one-to-one
correspondence between the sets d and f!J (i.e., to every a E d there corre-
sponds exactly one bE f!J and vice versa: a - b), which preserves the alge-
braic operations.
Two linear scalar product spaces and <1>' are, thus, isomorphic if from
 3 f - I' E <1>',
(2.8)
 3 9 - g' E <1>'
it follows that
(Xf + pg - (XI' + pg', (2.9)
and
(f, g) = (f', g')" (2.10)
Scalar product spaces (and in particular Hilbert spaces) for which (2.9)
and (2.10) are fulfilled are also called isometric, It often happens that two
scalar product spaces are isomorphic as linear spaces, i.e., are in a one-to-one
correspondence which fulfills (2.9), but are not isomorphic as scalar product
spaces, i.e., the correspondence does not fulfill (2.10).
1.3 Linear Operators

Vectors in 1R3 can be transformed into each other. One example is the
rotation R of a vector a into a vector Ra = b. In analogy to this one defines
transformations or linear operators on a linear space <1>. A function A,
A: --+ <1>, that maps each vector into a vector 1/1 E <1>,
A, IjJ E and a E C it fulfills the conditions
A(</> + 1jJ) = A</> + AIjJ, (3.1)
A(a</» = aA</>. (3.2)
It is called antilinear if it fulfills instead of (3.2) the relation
A(a</» = a* A</>, (3.3)
where a* is the complex conjugate of a.
The operations of addition of two operators A + B, multiplication of an
operator by a complex number aA, and multiplication of two operators AB,
are defined in the following way:
(A + B)</> = A</> + B</>, (aA)</> = a(A</», (AB)</> = A(B</», (3.4)
for all </> E <1>. It is easily verified that A + B, aA and AB are linear operators
if A and B are linear operators.
Of special interest are the zero operator, denoted 0, and the unit operator
or identity operator, denoted f, which are defined by
01jJ = 0, fljJ = IjJ (3.5)
for every IjJ E <1>. Note that 0 on the left is the zero operator, while 0 on the
right is the zero vector of (2.1c).
For every linear operator A defined on all of <1>, one can define an operator
A t by (A t </>, 1jJ) = (</>, AIjJ) for every </>, IjJ E <1>. The operator A t is called the
adjoint operator of A. An operator for which At = A is called self-adjoint or
Hermitian.!
The definition of linear operators was inspired by the properties of
transformations on the three-dimensional space. Linear operators on a
linear space may represent such transformations on the physical three-
dimensional space, but they can also have other physical interpretations. In
particular, in quantum physics they represent physical observables.
An important notion is the notion of eigenvalue and eigenvector. A
symmetric tensor 0 in three dimensions can be diagonalized by transforming
to its principal axis:
(3.6a)
where I(a) is the eigenvalue and a the corresponding eigenvector. Similarly
one defines eigenvalues and eigenvectors in a linear space <1>. A nonzero
vector IjJ E is called an eigenvector of the linear operator A if
AIjJ = AIjJ with A E C. (3.6b)
A is called the eigenvalue of A corresponding to the eigenvector 1jJ. For a
given operator A there may be many (perhaps infinitely many) different
1 We will usually use the term Hermitian if we do not want to distinguish between the mathe-
matically precisely defined notions self-adjoint, essentially self-adjoint, and symmetric.
eigenvectors with different eigenvalues. There are also linear operators on

linear spaces which do not have one single eigenvector in that space.
If A is a Hermitian operator, At = A, then eigenvectors and eigenvalues
have the following properties:
(1) All eigenvalues are real.
(2) If </>1 and </>2 are eigenvectors of A with eigenvalues Al and A2 ,
respectively, and if Al ¥ A2 , then </>1 and </>2 are orthogonal to each
other, (</>1' </>2) = O.
The notion of eigenvalue is important in quantum physics: As mentioned
above, the operators represent the observables of a physical system. The
eigenvalues then represent the numbers which are obtained in a measurement
of one of these observables.
An operator B is called the inverse of an operator A if
BA = AB = I. (3.7a)
It is denoted by
B=A- 1 • (3.7b)
A linear operator U is called a unitary operator if
UtU = UU t = I. (3.8a)
Because of (3.7) one defines a unitary operator also by the condition:
U t = U- 1 • (3.8b)
Let A and B be two operators. A and B are said to commute if
[A,B] == AB - BA = O. (3.9)
[A, B] is called the commutator of A and B.
A set d is an (associative) algebra with unit element iff
(a) d is a linear space.
(b) For every pair A, BEd, there is defined a product AB E d such that
(AB)C = A(BC), (3. lOa)
A(B + C) = AB + AC, (3.10b)
(A + B)C = AC + BC, (3.10c)
(aA)B = A(aB) = aAB. (3.10d)
(c) There exists an element lEd such that
IA = AI = A (3.11)
for all A Ed.
A subset d 1 of an algebra is called a subalgebra of d if d 1 is an algebra with
the same definitions of the operations of addition, multiplication by a
number, and multiplication as given for d; i.e., iffrom A, B E d 1 , it follows
that A + BE d 1 , aA E d 1 , and AB E d 1 .
An algebra.s;1 is called a *-algebra if we have on the algebra a *-operation

(involution) A HAt that has the following defining properties:
(d) (aA + bB)t = aAt + 5Bt,
(AB)t = Bt At,
(3.12)
(At)t = A,
It = I,
where A, B E .s;1 and a, b E C. From the definitions of the sum and product
of operators with a number-given in (3.4)-and from the definition of the
formal adjoint operator, one can see that the set oflinear operators fulfills all
the axioms (a), (b), (c), and (d) ofa *-algebra. Thus the set of linear operators
in a linear space forms a *-algebra. A *-subalgebra of this algebra is called
an operator *-algebra. It can be shown that in a certain sense every *-algebra
can be realized as an operator *-algebra in a scalar-product space (generaliza-
tion of the Gelfand-Naimark-Segal reconstruction theorem). In quantum
mechanics physical systems are assumed to be described by operator
*-algebras.
A set Xl' X 2' . . . , X n of elements of .s;1 is called a set of generators, and .s;1
is said to be generated by the Xi (i = 1, 2, ... , n) iff each element of.s;1 can be
written
n n
A = cI + '"
L. ciX. + '"
L. cijX.X.J + ... '
1 1
(3.13)
i= 1 i.i
where c, d, cij, ... E C.
Defining algebraic relations are relations among the generators
P(X) = 0, (3.14)
where P(X i) is a polynomial with complex coefficients of the n variables Xi'
An element B E .s;1,
(3.15)
where b, bi, ... E C, is equal to the element A iff (3.15) can be brought into
the same form (3.13) with the same coefficients c, ci, cij, ... by the use of the
defining relations (3.14).
1.4 Basis Systems and Eigenvector Decomposition

As in the three-dimensional space ~3, one can also introduce a system of
basis vectors in a general linear space <1>. In ~3 one conveniently chooses a
system of three normalized vectors
Ileill = 1,
which are orthogonal to each other
i,j=1,2,3.
I.4 Basis Systems and Eigenvector Decomposition 9
Such an orthonormal basis system can also always be chosen in a linear scalar
product space <1>. We denote these basis vectors in various ways by
i = 1, 2, 3, ... , N. (4.1)
Their scalar products are written in one of the following ways:
i,j = 1,2, .... (4.2)
If is N-dimensional then there are N linearly independent vectors in this

orthonormal basis system. N can be infinite.
The basis system fulfilling (4.2) can be chosen arbitrarily, but it is con-
venient to choose it such that the particular physical problem under in-
vestigation takes its simplest mathematical form. For example, if one
describes a rigid body with moment of inertia tensor ~ in the usual three-
dimensional space, then it is useful to choose the basis system, ei , i = 1,2, 3
such that
i.e., choose-according to (3.6a)-eigenvectors of the tensor L

Similarly, it is extremely useful to choose as the basis system for the space
of physical states, <1>, eigenvectors of an operator A which represents an
important observable of the physical system under investigation (most
frequently one chooses for A the energy operator H, the position operator
Q or the momentum operator P). Thus one would want to choose for the
basis system a set of vectors ei which fulfill
(4.3)
where the a i are the eigenvalues. These normalized eigenvectors are often
denoted by
The symbol I >with a letter <jJ, a or i in it denoting a vector <jJ, an eigenvalue

a or an index i of an eigenvalue is called a ket; the symbol <I with a letter in
it is called a bra. 2
In the three-dimensional space [R3 every vector v can be expanded with
respect to a basis system of eigenvectors of any symmetric tensor, i.e.,
(4.4a)
are the coordinates or components of the vector v with respect to the basis
of eigenvectors {e i } of the symmetric tensor ~ (3.6a). The same can be proven
2 P. A. M. Dirac (1958).
IO I Mathematical Preliminaries
for any finite-dimensional linear scalar product space <1>. This means that:
For every Hermitian operator A in a finite-dimensional space <1> there exists
a system of eigenvectors
i = 1, 2, ... , N = finite, (4.3)
such that every vector ¢ E <1> can be written as
N N
¢ = Z>i(e i , ¢) == Z>i(ail¢), (4.4b)
i~ 1 i~ 1
where the complex numbers
Ci = (e i , ¢) == (ail¢) (4.5)
are the components of the vector ¢ with respect to the basis {eJ
The set of a/s (which are real if A is Hermitian) is called the spectrum of
A and the above statement is called the spectral theorem for the operator A
in a finite-dimensional space. Equation (4.4b) is called the spectral decom-
position of the vector ¢ or the eigenvector expansion of ¢.
For infinite-dimensional spaces the above statement is in general not
correct. Furthermore, quantum physics also requires operators whose set of
eigenvalues is a continuous set, or even a union of a discrete set and a con-
tinuous set.
H could have been that all operators which appear in quantum physics
have the property that the set of their eigenvalues is discrete. That would
have been the case if the measurement for every observable in quantum
physics could only lead to a discrete set of numbers. Then only an infinite-
dimensional generalization of (4.3) and (4.4) would be needed. However,
there are observables in physics whose measurement can lead to any number
out of a continuous set of numbers (e.g., the observables momentum and
position can in many cases take any value x with - 00 < x < + 00). There-
fore we need not only the infinite-dimensional generalization (which can
be obtained in the Hilbert space) but also the continuous, infinite-dimensional
generalization of (4.3) and (4.4). There exist in fact spaces <1>-or more
precisely, there exist topologies for infinite-dimensional linear spaces-for
which the generalization of the finite-dimensional spectral decomposition
can be proven for all self-adjoint operators needed in physics. 3 This gener-
alization is the nuclear spectral theorem. As the eigenvector decomposition
is so important for physics we will only use spaces for which this theorem
is valid.
We cannot present the mathematics here and will explain the discrete and
continuous eigenvector decomposition in an infinite-dimensional space in
analogy to the finite-dimensional case (4.4b) or (4.4a). We consider the
discrete and continuous cases separately; the general case of an arbitrary
self-adjoint operator of physics will be a combination of these two cases.
3 These spaces and their conjugates <l>X (cf. Section I.7 below) (and the closely related
theory of distributions) and the triplet c Yf c x they form with the Hilbert space Yf did
in fact not exist when they were needed for quantum mechanics. The creation of these mathe-
matical structures was inspired by the development of quantum theory.
We will call the self-adjoint operator with a discrete set of eigenvalues H

and the operator with a continuous set of eigenvalues Q. Then the spectral
theorem asserts:
There exists a system of eigenvectors IEJ in the discrete case and Ix)
in the continuous case;
HIEn) = En lEn), En = E o ,E 1 ,E 2 , · · · , (4.3d)
Qlx) = xix); - 00 ~ m ~ x ~ M ~ + 00, (4.3c)
such that every ¢ E can be expanded in terms of these eigenvectors:
00
f:
(4.4d)
n=O
¢ = dxlx)<xl¢)' (4.4c)
and ¢ = 0 if and only if all its components are zero, i.e., (En I¢) = 0 for all
En and <xl¢) = 0 for all X.4
A system of eigenvectors lEn) or Ix) with these properties is called com-
plete or a basis system. Thus the spectral theorem asserts the existence of a
complete system of eigenvectors of a self-adjoint operator. (En I ¢) or <x I ¢)
are called the coordinates or components of ¢ with respect to the basis
system {I En)} or {I x)}, respectively. They are, as in the three-dimensional
case, the scalar products of the eigenvectors with ¢:
<x I ¢) = (Ix), ¢), (4.5c)
(Enl¢) = (lEn)' ¢). (4.5d)
Thus (En I¢) is the infinite-dimensional generalization of the Vi in (4.4a),
and <x I¢) is the continuous infinite-dimensional generalization of Vi'
Whereas the lEn) are proper eigenvectors, the Ix) are called generalized
eigenvectors, or, eigenkets. Though we can manipulate them as if they were
proper eigenvectors, mathematically there is an important difference between
the discrete basis vectors I En) and the continuous basis vectors Ix): the IEn)
are in while the Ix) are in x, the space of continuous antilinear functionals
over <1>. We shall define and explain these mathematical notions in Section
1.7. Here we shall try to convey the meaning of these generalized eigenvectors
by analogy to the finite-dimensional case.
The set of eigenvalues En in (4.3d) is called the spectrum of the operator
H. If H has a discrete set of eigenvalues, the spectrum is called discrete. All
the corresponding eigenvectors IEn) enter into the discrete basis vector
expansion (4.4d) and there are no further eigenvectors with discrete eigen-
value that enter into the basis vector expansion (4.4d). The set of continuous
eigenvalues x, whose eigenvectors enter into the generalized eigenvector
expansion (4.4c), is called the continuous spectrum of the operator Q.
4 The simple non degenerate form (4.4d), (4.4c) is valid if the operator A (H or Q) is cyclic,
i.e., if there exists an f E ct> such that A"f == 1(n) generate the entire space ct>, which means that
any cP E ct> can be written as cp =In 1(n)c(n) where c(n) are complex numbers. Degenerate spectra,
which occur when more than one quantum number is needed, will be discussed later in the text.
In general-and this depends upon the properties of the space <1>- there are
more generalized eigenvectors of Q-i.e., kets which fulfill (4.3c) and whose
precise definition will be given in Section 1.7 -than enter in the eigenvector
expansion (4.4c). Their generalized eigenvalues we will not include in the
definition of the continuous spectrum. Whereas the discrete eigenvalues of a
self-adjoint operator are always real, the generalized eigenvalues need not
be real; they can be real or complex, and even if they are real they need not
necessarily belong to the spectrum, i.e., appear in the integral (4.4c). But
for a self-adjoint operator there is always a real subset of the set of generalized
eigenvalues whose eigenvectors are complete.
The most general form of the spectral theorem for an operator A
representing a physical observable is a combination of (4.4d) and (4.4c):
(4.4g)
where the sum is over the discrete spectrum and the integral is over the
(absolutely) continuous spectrum of A. It can happen that some or all values
a i appearing in the sum also appear in the integral. Then they are called
discrete eigenvalues in the continuous spectrum. If this happens for ak then
la k ), is still orthogonal to allla) including la k ), i.e.,
(a k I4/) = 0 for 4/ = J
da la)<aI4/)·
The spectral theorem (4.4b), (4.4d), (4.4c), and (4.4g)-which we have to

accept here without proofs-is the central statement, from which the other
results of this section follow.
To see that the coordinates (En Icf» are indeed what their notation indicates,
namely the scalar product ofthe vector with the basis vector IEn), we calculate
the scalar product of (4.4d) with the eigenvector IEm)' From (4.4d), it follows
that
00
(IE m), cf» = L (IE m), IEn»(En Icf». (4.6d)

n=O
Since IEm) and IEn) are eigenvectors of the same Hermitian operator H,
if En #- Em they must be orthogonal to each other,
(4.7')
For En = Em we normalize them:
(4.7")
5 The finite-dimensional case (4.4b) is easily reduced to the problem of the number of roots
of a polynomial of degree N, as shown in Section I.5 below in particular by Equation (5.17).
Equations (4.4d) and (4.4c) need much more mathematical preparation, cf. I. M. Gelfand et al.
(1964), Vol. 4, or K. Maurin (1968).
This we combine and write

(IE m), lEn» = (Em lEn) = bEnErn = bnm ; n, m = 1,2, ... , (4.7d)
where the Kronecker b is defined by
b _ b _ {1 for n = m; En = Em'
nm - EnErn - 0 for n "# m; En "# Em' (4.8)
So these eigenvectors of the self-adjoint operator H have the property (4.2)
as required of orthonormal basis vectors. Inserting (4.7d) into (4.6d) one
obtains:
00
(IEm), <1» = L bmiEn I<1» = (Em I<1». (4.5)
This is the expected identity (4.5d).

The spectral theorem (4.4d) can be written in different forms: one can
omit the arbitrary vector E <D on both sides of (4.4d) and obtain the
spectral resolution of the identity operator l:t
(4.9d)
One can multiply both sides of (4.4d) with the operator H,

00 00
H = L HIEn)(Enl<l» = L En IEn)(En I <1»,
and then omit the arbitrary vector on both sides to obtain:
00
H = L En IEn)(En I· (4.lOd)
This identity between the operator H and the weighted sum of operators
IEn)(En I is called the spectral resolution of the self-adjoint operator H with
discrete spectrum.
One can take the scalar product of (4.4d) with another t/J E <D, then one
obtains
00 00
(t/J, <1» = L (t/J IEn)(En I<1» = L (En It/J)*(Enl <1». (4.11d)
In particular if one chooses t/J = one obtains

00 00
11<1>112 = (, <1» = L ( IEn)(En I<1» = L I(En I<1»1 2. (4. 12d)
Equation (4.11d) is the analogue of the formula

3 3
v' X = L vie j ' ejxj = L VjXj
i,j=1 i=1
for the ordinary scalar product in (R3.
t The quantities An = IEn)(En I are projection operators as defined in the mathematical insert
on p. 58.
As in the three-dimensional space 1R 3 , a vector ¢ is completely specified

by its components (En I¢) with respect to a given basis IEn)' But unlike the
three-dimensional space IR\ where any sequence of three real numbers
(Xl' Xl' x 3 ) defines a vector x, or unlike the N-dimensional complex space
eN where any sequence of N complex numbers (¢1, ¢l, ... , ¢N) defines a
vector x, an arbitrary infinite sequence of complex numbers
does not in general define a vector in <1>. As can be seen from (4. 12d), in order
that the scalar product be defined, the infinite sequence that defines a vector
must at least fulfill the condition
00
LI¢il l < 00, (4.12d')

i= 1
i.e., it must be square summable. If one further wants to demand that every
operator A of the set of operators which represents physical observables be
defined in the whole space <1>, one has to require that all A¢ are well defined
in <1>, which means that (A¢, A¢) must be finite. Choosing A = HP where
p = 0, 1, 2, ... is any power, one obtains for this requirement the following
condition:
00
(HP¢, HP¢) = L (¢IHPIEn)(EnIHPI¢)

n=O
= LE;PI(Enl¢)ll<oo forany p=0,1,2, .... (4.1 3d)

n=O
Thus not only will {(En I ¢) In = 0, 1,2, ... } have to be square summable,
but also {£P(Enl¢)} has to be a square sum mabie sequence for any p = 0,
1,2, ....
Fortunately these (topological) questions of what happens at infinity are
not very relevant for physics as only a finite number of the I(En I¢) I can be
determined experimentally.
We now turn to the continuous spectrum and repeat the above con-
siderations for the continuous case. We calculate the scalar product of ¢
with the generalized eigenvector IX)6 using Equation (4.4c):
(Ix), ¢) == fdY(,x), ly»<yl¢)·
This we rewrite with the following definition of the new symbol

<xly) == (Ix), Iy»
6 More precisely, using the notions that will be introduced in Section 1.7, we should say that
we calculate the value (I x), 4» of the functional Ix) at the vector 4> E <1>, (I x), 4» = <x 14» is a
generalization of the usual scalar product.
as
(Ix>, ¢) = f dy <xly><yl¢>· (4.6c)
<x I¢ >is the coordinate of the vector ¢ along the direction of the basis vector
Ix). (Ix>, ¢) is the scalar product of ¢ with the basis vector Ix>. These two
quantities should be the same, as stated by (4.5). Therefore (4.6c) has the form
<xl¢> = Joo dy <xly><yl¢>. (4.6c')

-00
The scalar products <y I¢ >are functions of the continuous variable y in

the same way as the scalar products (En I¢) are functions of the discrete
variable En. Equation (4.6c') therefore says that the mathematical quantity
<x Iy> has the property that it maps the function ¢(y) = <y I¢ >by integration
into its value at the point x: ¢(x) = <x I¢>. There exist no well-behaved and
not even a locally integrable function which has this property.
Such a quantity is called a distribution or a generalized function. The
distribution <xly> defined by (4.6c') for a class of well-behaved functions
¢(x) = <x I ¢ >is called the Dirac b-function and is denoted in analogy to
(4.7d)as
<xly> = b(x - y). (4.7c)
b(x - y) is a generalization of the Kronecker bEnEm which is usually defined
by (4.8) but which could as well have been defined by
00
(Ernl¢) = L bEmdEnl¢) (4.6d')

n=O
for a class of infinite sequences {(En I¢)}. When we will use function se-
quences of the b-type in Section 11.8, we will see in which sense b(x - y) can
be considered as a generalization of the right-hand side of (4.8).
The eigenvectors lEn) are normalized to 1 by (4.7"); the generalized
eigenvectors Ix> fulfilling (4.7c) are called function normalized. They are
not dimensionless, but have the dimension l/Jdim dx. For example, if dx
has the dimension cm, then <x' Ix> has the dimension cm -1, and Ix> has
the dimension cm- I/Z •
Instead of the generalized eigenvectors with b-function normalization
(4.7c) one could also choose generalized eigenvectors of Q with a different
normalization. Instead of (4.4c) one writes
where the Ix}!, are again eigenvectors of Q:

Qlx}p = xlx}p'
and p(x) = dJ.1(x)/dx is a real nonnegative and integrable function.
In order that the new components of fjJ, the p{x I fjJ), be the scalar product
of fjJ with the new eigenvectors Ix}p' i.e., in order that
p{xlfjJ) = f p(y) dy p{xIY}pp{ylfjJ),
one has to demand

dJ1(Y) p{xly}p = dy <xly) = dy l5(x - y).
So the normalization of the new generalized eigenvectors is
p{xIY}p = e~~») -ll5(X - y) = p-l(y)t5(X - y).
Thus when the integration contains the weight function p(x), the generalized
eigenvector normalization contains the factor p-l(X).
The transformation between the two basis systems is
1
Ix}p = Ix) 0::\' (4.14)
v p(x)
The most appropriate choice for p(x) depends upon the property of the
operator Q and its relation to the other operators of the problem. 7
As in the case of the discrete spectrum the spectral theorem (4.4c) can be
written in different forms. Omitting the arbitrary vector fjJ one obtains the
resolution of the identity:
1= fdX,X)<X' = fp(X)dX'X}pp{X I. (4.9c)
Multiplying both sides of (4.4c) with the operator Q and then omitting the
arbitrary vector fjJ one obtains after (4.3c) has been used
Q= f dx xlx)<xl (4.lOc)
which we call the spectral resolution of the self-adjoint operator Q with

(absolutely7) continuous spectrum.
The scalar product of two vectors fjJ, '" E <1>, obtained from (4.4c), is
("', fjJ) = f dx <"'lx)<xlfjJ)· (4. 11 c)
7 The nuclear spectral theorem for an arbitrary self-adjoint operator does in fact not assert
(4.4g) but (4.4c p) with a general measure dJl.(x), and it does not say anything about the spectral
measure dJl.(x) in addition to the assertion of its existence. However, all operators used in physics
are of the special kind for which either dJl.(x) = p(x) dx with p(x) as described above (such
operators are said to have an absolutely continuous spectrum and for them one can always
make the transformation (4.14) from Ix}p to Ix», or they have the property that dJl.(x) =
Lx,o(x - x,) dx (these are the operators with discrete spectrum), or they have both an ab-
solutely continuous and a discrete spectrum. So (4.4g) is the most general form needed for
physics.
J
From (t/J, ¢) = (¢, t/J)* and from (¢, t/J) = dx <¢Ix)<xlt/J) for any
¢, t/J E one concludes that for the generalized scalar product one has the
same relation as for the scalar product:
<t/Jlx) = <xlt/J)*·
Using the notation
<xl¢) = ¢(x), <xlt/J)* = t/J*(x),
one can write (4.11c) in the form
(t/J, ¢) = f dx t/J*(x)¢(x). (4.llc')
In particular, if one chooses t/J = ¢ one obtains
11¢11 2 = (¢, ¢) = f dx ¢*(x)¢(x) = f dx 1¢(xW. (4.l2c)
From this we see that not any arbitrary function ¢(x) can give the components
of a vector ¢ E with respect to the continuous basis system Ix), but only
those functions for which the integral on the right-hand side of Equation
(4. 12c) exists, i.e., the square integrable functions. s
If one demands of the space that on all its elements the operator Q and
any arbitrary power thereof, QP (p = 0, 1, 2, ... ), be well defined, then one
must have that
(4.l3c)
Thus I¢(x) I must decrease faster than any power of x. If other operators are
also to be defined everywhere in further conditions will have to be imposed
on the components <x I¢) of ¢ E <1>. Thus the realization of must be much
better than L 2 . An example of a realization of is the Schwartz-space S. S is
defined as the space of infinitely differentiable complex-valued functions
which together with their derivatives vanish at infinitely more rapidly than
any power of 1/x. We call those functions well behaved.
It can be that the space is such that the components of all vectors t/J E 
with respect to the continuous basis Ix), t/J(x) = <x I t/J), are boundary values
of analytic functions t/J(z) on the complex plane or a domain of the complex
8 The scalar product space in which the scalar product is realized by the integral (4.11c') is
called the space of square integrable functions L 2. If the scalar product space is complete then
it is called a Hilbert space. For this reason a scalar product space is also called a pre-Hilbert
space. The space of square (Lebesgue) integrable functions is one realization of the Hilbert
space. Another realization of the Hilbert space is the space of square summable infinite se-
quences, cf. (4.12d').
plane, e.g., the lower half plane. Then the contour of integration in (4.11c')
can be deformed and one obtains
(ljJ, ¢) = L+oooo dx ljJ*(x)¢(x) = L dz ljJ*(z)¢(z)
= L dz (ljJlz)(zl¢)· (4.1S)
C(}can be any contour which is obtained from the contour along the real
axis by deforming it into the domain of analyticity without passing over a
singularity. This defines the generalized eigenvectors Iz) of the self-adjoint
operator Q:
Qlz) = zlz) (4.16)
with complex eigenvalue z. These generalized eigenvectors Iz) can also be
used in a generalized basis vector expansion. This is obtained from (4.1S) by
omitting the arbitrary vector ljJ E 
¢ = L dz Iz)(zl¢)· (4.4cc)
Thus instead of using the generalized basis vector expansion (4.4c) with
respect to the generalized eigenvectors Ix) with x E IR, one can in this case
as well use the generalized basis vector expansion (4.4cc) with respect to the
generalized eigenvectors Iz) with complex eigenvalues z E IC. This possibility
will be used in the description of decaying states.
1.5 Realizations of Operators and of Linear Spaces
We have already mentioned that the components of all vectors ¢ E with
respect to a basis system constitute a realization of the space <1>. A three-
dimensional vector x in R3 is thus realized by the sequence (Xl' X2, X3) of its
components Xi = ei • x. The components, of course, depend upon the basis
system chosen. If one takes another basis system e; the components of the
same vector x change: x; = e; . x.
In a finite-dimensional linear scalar product space the components are
finite sequences of complex numbers (4.S). In an infinite-dimensional space
they are either infinite sequences ¢i = (Ei I ¢) == (i I¢) which fulfill certain
conditions such as (4.12d') and (4.13d), or continuous infinite sequences
¢(x) = (x I¢), i.e., functions of a continuous variable x, which also fulfill
additional conditions like (4.12c) and (4.13c).
As in the three-dimensional case, if one changes the basis system of the
space, the components of a given vector ¢ change too. Thus, if we take in
addition to H of (4.3d) another operator A which has, like H, a discrete
spectrum
Ala) = aila), (S.l)
I.5 Realizations of Operators and of Linear Spaces 19
then the same vector cp of (4.4d) can also be expanded as

<Xl
cp = L la)<a;lcp) = L li)<ilcp), (5.2)

;=0
where we defined for writing convenience: Ii) la). =

If A and H do not commute, then the components <a;lcp) and (En Icp) are
entirely different. The components can be written as a column matrix
(5.3)
(5.4)
Column matrices can be added, by adding each of their components, they

can be multiplied by a number by multiplying each component and a scalar
product between column matrices can be defined as by the expression furthest
to the right in Equation (4.11d). With these definitions it is easy to see that
the set of column matrices: <ilcp), <ill/l), <ilx), ... , i = 1,2,3, ... form a
linear scalar product space Cm(cf. also Problem 1). In the same way the set
of column matrices of components with respect to the basis system
{IEn)iEn = E1 , E2 , · · .}, the (Enlcp), (En I1/1), (En' X), ... , form a linear scalar
product space C~"). It is easy to see that C~") and C are isomorphic to each m
other and isomorphic to the space «D:
«ilcp» +-> ((Enlcp» +-> cp,
«i I1/1» +-> ((En I1/1» +-> 1/1.
Therewith we can define the space «D by specifying the set of all infinite
column matrices. For instance we shall define C~") as the space of all (En Icp)
which fulfill (4.13d) with En = (n + 1-), n = 0, 1,2, .... «D can then be defined
as the space isomorphic to it. 9 Thus the same vector can be represented by
entirely different column matrices.
When vectors are realized by column matrices, operators are realized by
quadratic matrices. We obtain these matrices of an operator B in the following
way:
Calculate
Bcp = L Bli)<ilcp), (5.5)
;
9 C~n) has a natural topological structure whicQ then inherits. is then the largest space
in which all the operators HP, p = 0, 1,2, ... are continuous operators, and its topological struc-
ture could have been defined by this requirement.
and take the scalar product of this equation with Iaj )

0IB¢) = L 0IBli)<il¢)· (5.6)
This can be written in matrix notation as
<IIB¢») (lIBll) <IIBI2) (1IBI3) ... )(ll¢»)

(
<2IB¢) _ <2IBll) <2IBI2) <2IBI3) <21¢)
(5.7)
(3IB¢) - (3IBll) <3IBI2) ... ... (31¢)'
·· .. ...
· .
The right-hand side of (5.6) gives the definition of multiplication of a quad-
ratic matrix with the column matrix on the right-hand side of (5.7). Thus we
see that the linear operator B is realized by the infinite matrix
lIBll) <IIBI2) ... )
B+-+(jIBli) = ( (2IBll) (2IBI2) ... , (5.8)
(3IBI1) ...
which is called the matrix of the operator B with respect to the basis system
{Ii)!i = 1,2,3, ... }.
Now, let D be a second operator and apply D + B to the basis vector Ii)
using (3.4)
(B + D)li) = Bli) + Dli).
Taking the scalar product of both sides of this equation with the basis vector
Ij) one obtains (using (2.2d»:
01(B + D)li) = 01Bli) + 0IDli). (5.9)
Thus the matrix of the sum of two linear operators is equal to the sum
of the matrices, as defined by the right-hand side of (5.9). Let us now con-
sider the matrix element of DB and use the resolution of the identity (4.9d)
for the basis system (5.1):
00
0IDBli) = L 0IDln)<nIBli). (5.10)

n=Q
Written in matrix notation this has the form:

<IIDBI1) (1IDBI2) <IIDBI3) ... )
( <2IDBll) (2IDBI2) (2IDBI3) .. .
....... .
lIBll) <IIBI2) ... )
_ (1IDI1) <IIDI2) <IIDI3) :::) (2IBll)
- (2IDI1) <2IDI2) <2IDI3)
....... .
(3IBll) « (2IBI2)
(3IBI2)
.. .
.. .
(5.10')
I.5 Realizations of Operators and of Linear Spaces 21
where the right-hand side of (5.10) gives the definition of the multiplication
of two (infinite-dimensional) quadratic matrices.
Thus in the realization of the space by the space of column vectors, the
operators are realized by matrices in such a way that the sum of two operators
corresponds to the sum of their matrices and the product of two operators
corresponds to the product of their matrices.
If the basis system is chosen to be a system of eigenvectors of the operator
A as in (5.1), then the matrix of A with respect to this basis system is diagonal
as from (5.1) follows
(5.1 I)
or, written as a matrix,
o 0 ... )
o 0 .... (5.11')
o 0 .. .
The matrix elements with respect to the basis system of eigenvectors of A

of all other operators that do not commute with A have nonzero off-diagonal
matrix elements.
As the proper eigenvectors of H are also elements of one can use (4.4d)
and expand them with respect to the basis system Ii) of (5.1) and vice versa
00
lEn) = L li)<iIEn); (5.12)
00
Ii) = L IEn)(Enl i). (5.12')
Taking the scalar product ofthis equation with ¢ and then taking the complex
conjugate one obtains
00
(Enl¢) = L(Enli)<il¢). (5.13)
Thus (En Ii) are the matrix elements of a matrix that transforms the infinite-
column matrix (5.4) into the infinite-column matrix (5.3). It is called the
transition matrix (or transformation matrix) between the two basis systems
{I i) Ii = 1, 2, ... and {I En) I n = 1, 2, ... }. Its elements (En Ii) are also called
transition coefficients. Taking the scalar product of (5.12') with U) gives
00
<J Ii) = (Jij = L <J IEn)(En Ii) = L (En Jj)*(E nIi). (5.14')
n = 1. 2 •...
Similarly one obtains

00
(Em lEn) = (Jmn = L (Emlj)<JIEn) = L (Em Ij)(E nIj)*· (5.14)

j= 1
A matrix whose matrix elements fulfill the condition (S.14) is called a

unitary matrix. If the matrix elements are also real, which happens sometimes
in <l> (and always in a real linear space), then the matrix is called orthogonal.
If the matrix elements of the operator A in the basis IEn) are given,
(En IA IEm), then one can obtain the matrix elements <i IA Ij) by using the
transformation (S.12'):
00 00
aAj = <iIAlj) = L L <iIEn)(EnIAIEm)(EmID· ( S.1S)

n=1 m=1
One says, the transition matrix transforms the matrix (En IA IEm) into its
diagonal form. Multiplying both sides of (S.lS) by <Erli), summing over i
and using (S.14') one obtains:
00
L (ErIAIEm)(Emlj) = a/Erlj). (S.16)

m=1
If the eigenvalues a j are unknown, then this is the eigenvalue problem in

matrix notation. For the case that the space is N-dimensional (instead of
co-dimensional) this is a system of N linear homogeneous equations lO for
the N unknown (Emlj):
N
L [(En IA IEm) - ac5 nm J(Em Ij) = O. ( S.17)
m=1
It has nonzero solutions for (Em U) if and only if
(S.18)
The eigenvalues of the matrix (En IA IEm) are the N solutions of this
equation. These solutions are not necessarily distinct. Equation (S.18) is a
polynomial of degree N and, by a famous theorem of algebra,lo every
polynomial of degree N has N (in general complex but if A is Hermitian
then real) solutions.
We now turn to the case that one of the basis systems,
{lx)l-co < x < + co}, (S.19)
is continuous. Then the vector ¢ is realized by a function of a continuous

variable x, rather than by a function of a discrete variable as in (S.3) and (S.4):
(S.20)
As in the discrete case, the space <l> can be defined by its isomorphism
(S.20) to a space of functions. We want to consider the particular case where
<l> is the space of vectors whose components <x I¢) are elements of the func-
tion space S.
10 See, e.g., A Lichnerowicz (1967).

The eigenvector lEn) is a particular vector of the space <1>; so one can use
the continuous basis system expansion (4.4c) for it:
(5.21)
The <x IEn) are the analogue of the transition matrix elements are functions of the discrete variable En
and for fixed value of En they are functions of the continuous variable x.
They also occur if one takes the scalar product of the basis vector expansion
(4.4d) with the continuous basis vector I x> :
00
<xl</» = L <x IEn)(En I</». (5.22)

n=O
The <x IEn) constitute a particular set of functions in the space S, which, as
a consequence of (4.3d), have the property:
(5.23)
Because of this property they are called eigenfunctions of the operator H.

Equation (5.22), which is an immediate consequence of the eigenvector
expansion (4.4d), is called the eigenfunction expansion of the function
<xl</» E S.
We want to illustrate this on a well-known example. We choose for the
space of functions the subspace K(a) c S of all functions </>(x) = <x I</»
which are identically zero outside the domain Ix I < a.
Let the operator Q have the continuous spectrum {x I - a < x < + a}. So
the spectral representation of an arbitrary vector </> E is
</> = L+: dx Ix)<x I </> >. (5.24)
Let the self-adjoint operator H = p2 be defined by

d2
(xIHI</» = - dx 2 (xl</» (5.25)
for every component <x I</» of any </> E <1>. In order that Q, P and any power
of these operators be defined in <1>, the space of components <x I</» must be
the space of continuous infinitely differentiable functions for which
f _ a dx
a 1
Xn
dm (X I</» 12 <
dxm CIJ.
Further
<x?: al</» = <x :::;; -al</» = O. (5.26)
(This boundary condition is important to establish the self-adjointness of H,

cf. Problem 14.) Let the eigenvector of H be denoted by In):
(5.27)
We want to find all eigenvalues En and the transition coefficients <x In).
From (5.25) and (5.27) follows
d2
- dx 2 <xln) = E.(xln). (5.28)
The solutions of this differential equation which fulfill the boundary condi-
tion (5.26) are
<xln) =
1
Jacos 2a x (mc) for n = 1, 3, 5, 7, ... ,
(5.29)
1 . (nn ) for n = 2,4, 6, ....
<xln) = Ja sm 2a x
The eigenvalues are

(5.30)
The normalization factor 1/Ja in (5.29) has been chosen so that <nln) = 1,
which according to (5.24) is given by
<nln) = f+a
-a dx <nlx)<xln)
= ~1 f+a
-a dx cos (n1r)
2a x cos (n1r)
2a x = 1, (5.31e)
or
<nln) = ~ t+aadX sinG: x) sin(;: x) = 1. (5.310)
In the last equalities of (5.31e) and (5.310) properties of integrals over the
trigonometric functions have been used. For n' # n one should have as a
consequence of the orthogonality of eigenvectors of self-adjoint operators
<n'ln) = f:aadX <n'lx)<xln) = O. (5.32)
But this can also be obtained by inserting (5.29) into (5.32) and calculating
the integrals over the trigonometric functions. Because of (5.32), one says
that the eigenfunctions <n'lx) and <nix) are orthogonal.
The eigenfunction expansion (5.22) for this particular case is
<xl¢)=¢(x)= I
n=1,3,5
ancos(n21rx)+
a
I
n=2,4,6
bn sin(n21r x )
a
for Ixl<a,
¢(x) = 0 for Ixl ~ O. (5.33)

The coordinates of the vector ¢ are given by:
bn= <nl¢) = f-a+a dx <nlx)<xl¢)
= Ja1 f+a dx sin (nn)

-a 2a x ¢(x); n = 2,4,6 ...
and
an = <n I¢) = f+a

-a dx <nlx)<x I¢)
= J; f:aa dx cos(;: x )¢(x); n = 1,3,5, ....
Equation (5.33) is called a Fourier series representation of the arbitrary

function ¢(x) E K(a). The coordinates of the vector ¢ with respect to the
basis system In), an and bn , are called the Fourier coefficients of the function
¢(x) = <x I¢). Because (5.22) is just a generalization of a classical Fourier
series (5.33) one also calls (5.22) and even (4.4d) often a Fourier series
representation or Fourier expansion. In general the eigenfunctions <x IEn)
will not be trigonometric functions, however (5.32) must always be fulfilled,
i.e., the <xIEn) always form an orthogonal system of basis functions. Well-
known examples of orthogonal basis functions are the Hermite Polynomials,
the Legendre Polynomials, and the Laguerre Polynomials. In particular,
for the space S, when the spectrum of Q is the entire real line IR, the Hermite
polynomials are an appropriate choice of basis functions.
So far we have considered transition coefficients between two discrete
basis systems <i IEn), and the transition coefficients <x IEn) between a discrete
and a continuous basis system. Now we want to go one step further and
consider transition coefficients between two continuous basis systems. We
choose for <l> the space realized by S and consider the operator P defined by
(cf. Problem 7)
1 d
<xlPl¢)=-;-d <xl¢) forevery ¢E, i.e., <XI¢)ES. (5.34)
! x
We want to consider (5.34) also for the case that ¢ is an eigenvector of the
operator P (extending the definition of P if this eigenvector is not in <1»:
Pip) = pip)· (5.35)
Equation (5.34) then reads
1 d
<xIPlp) = p<xlp) = -;-d <xlp)· (5.36)
I x
It is well known that this differential equation has solutions for any complex
value of P given by
(5.37)
None of these solutions, however, is square integrable so that Ip) ¢ 

(cf. Problem 9).
We can expand any </l E with respect to the basis system Ip) of eigen-
vectors of P:
</l = 1Spec!r. P
dp Ip)<pl</l), (5.38)
where the integral has to extend over the spectrum of the operator P. We
want to determine the spectrum of P. It is contained in the set of generalized
eigenvalues which is identical with all of C. We take the scalar product l l of
(5.38) with Ix)
<xl</l) = r
Jspectr. P
dp <xlp)<pl</l) =
V 2n
~ JdP eixP<pl</l). (5.39)
If the p integration in (5.39) extends over the real line, - 00 < p < + 00, then
(5.39) says that </lex) = <x I</l) is the Fourier transform of the function
¢(p) = <plcp). We therefore want to give a brief review of the properties of
the Fourier transform. 12
Let </lex) be an element of the space S, then the (inverse) Fourier transform,t
which we denote by Fpl[</l(X)] == ¢(p), is defined by
(S.40a)
The Fourier transform has the following properties:

(a) F- 1 maps S onto itself; that means ¢(p) also belongs, as a function
of p, to the space S.
(b) The Fourier transform which we denote by F x[ ¢(p)] and which is
defined by
(S.40b)
also belongs to the space S if $(p) E S.

(c)
d)n </lex) =;;-:
( --:1 -d 1 J+ 0Cl
dp pneixp¢(p) = F x[pn¢(p)] (S.40c)
I X V 2n - 0Cl
for any power n = 0, 1, 2, 3, ... , which we also write as
11 Section 1.4. footnote 6.

12 For the proofs and details see I. M. Gelfand et al. (1964), Vol. 2, Chapter III, or E. 1.
Beltrami and M. R. Wohlers, Distributions and the Boundary Values of Analytic Functions.
Academic Press, New York, 1966.
t One sometimes calls inverse Fourier transform what we call Fourier transform here and
vice versa; we may do the same.
(d)
1 d)n -
(- ~- 4>(p) = -
1 f+ 00
dx x ne- 1XP 4>(x) = F; 1 [Xn4>(X)]
.
(S.40d)
I dp fo -00
for any power n = 0, 1, 2, 3, ... , which we also write as
(- ~ ~rF;l[4>(X)] = F;l[X n4>(X)],
(e)
f dx 1jJ*(x)4>(x) = f dp \iI*(p)4J(p), (S.40e)
where \iI(p), 4J(p) E S are the Fourier transforms of ljJ(x), 4>(x) E S.

Thus the Fourier transform F and its inverse F- 1 establish two reciprocal
isomorphisms between the Schwartz space functions Sx of the variable x
and the Schwartz space functions Sp of the variable p.
We can now return to the problem ofthe spectrum of P. As every <x 14» =
4>(x) E S can be written according to (S.40b) and (S.37) in the form (S.39)
with the integration extending over the real axis - 00 
with respect to the basis system of eigenvectors of P. The spectrum of P is
continuous and the Ip) are generalized eigenvectors (not elements of <D).
The transition coefficients between the generalized basis systems {I x)} and
{I p)} are given by (S.37) with p E IR and x E IR. For fixed value of x they are
functions of the continuous variable p, and for fixed value of p they are
functions of the continuous variable x. They satisfy the differential equation
(S.36), but they are not well-behaved functions, i.e., elements of the function
space S. Therefore they are called generalized eigenfunctions or distributions.
It is easily seen by exchanging x and p in the preceding arguments and using
(S.40a) in place of (S.40b) that
1 .
<pix) = - e- 1pX• (S.41)
fo
Consequently we have shown that for this special case of transition co-
efficients between two continuous basis systems the following relation holds:
<pix) = <xlp)*. (S.42)
The exponential function is just one example of transition coefficients
between two continuous basis systems. Other examples that we will meet in
the second part of the book are the spherical Bessel functions.
Finally, in this section we want to consider the realization of operators in
terms of a continuous basis system. This is again done in analogy to the
discrete case. To obtain the analogue of (5.5) we apply the operator B to the
continuous basis system expansion (4.4c)
B¢ = f dx Blx)<xl¢)· (5.43)
Taking the scalar product 13 with the basis vector Iy) E { Ix) 100 < x < oo}
we obtain
<YIB¢) = f dx <yIBlx)<xl¢) (5.44)
which is the continuous analogue of (5.6). This is an integral transform

transforming the function <x I¢) into the function <y I t/I) = <y IB¢). The
continuous analogue of the matrix element (5.8), <yIBlx), is called the
kernel of the integral transform. Choosing for B = A and for ¢ = Ii), the
eigenvector (5.1) of A we obtain from (5.44)
f dx <yIAlx)<xli) = ai<yli). (5.45)
This is the analogue of (5.16). Ifthe eigenvalues a i and eigenfunctions <y Ii)
are unknown this is a homogeneous integral equation for the determination
of these values.
1.6 Hermite Polynomials as an Example of

Orthonormal Basis Functions
We start with the operator Q realized by the multiplication operator and

the operator P realized by the differential operator (Equation (5.34)) on the
space of functions S:
<xIQI¢) = x<xl¢), -00 < x < +00, (6.1)

1 0
<xIPI¢) =~;-- <xl¢), <XI¢)ES. (6.2)
luX
Another operator H is defined by

H = t(P 2 + Q2), (6.3)
and its eigenvectors are denoted by In):
(6.4)
We denote the space in which Q, P, H and any element of the algebra
generated by them act by <1>. Thus ¢ E ~ <x I ¢) E S.14
Section 1.4, footnote 6.

13
One can also construct <l> in the following manner (cf. Chapter II): Start from the algebra
14
of operators P, Q, and H which fulfills in addition to (6.3) the relation PQ - QP = (1/0/, and
construct the space <l> as the largest space in which this algebra is given (represented) as an algebra
of (continuous) operators and in which there exists at least one eigenvector of H.
The discreteness of the spectrum of H, which is implied by (6.4), is in fact

a consequence of the assumption that the eigenvectors In) of H lie in the
space or that (x In) E S, as will be shown below.
The transition coefficients between the Q-eigenvectors and the H-eigen-
vectors, (x In), are orthonormal basis functions of the space S in the sense of
(5.22). We want to determine (x In) and the eigenvalues E. explicitly.
We denote
)., = 2E •. (6.5)
Using (6.3) in (6.4) we obtain with (6.2) and (6.1):
2(xIHln) = (- ::2 + x 2 )(x ln) = A(xln). (6.6)
Thus we have to find the solutions

I/I(x) = (xln) (6.7)
of the differential equation
I/I"(x) + ()., - x 2 )I/I(x) = O. (6.8)
In addition we will demand that the I/I(x) E S. Then they must in particular
fulfill the boundary conditions:
I/I(x) --+ 0 faster than any polynomial of 1/x as x --+ ± 00 (6.9)
Equation (6.8) is highly singular at Ix I --+ 00 and for large values of Ix I

is given approximately by 1/1" - x 2 1/1 = 0, which has the solutions 1/1 '" e±x2 /2.
One therefore makes the ansatz
I/I(x) = c 1 e- x2 / 2 y(x) + c2 e x2 / 2 Yix) (6.10)
with Y2 and y expanded as power series in x:
(6.11)
The second term in (6.10) increases rapidly and cannot lead to an dement of
S, therefore C 2 must be zero. Inserting
I/I(x) = c 1 e- x2 / 2 y(x) (6.12)
into (6.8) one obtains the differential equation
y" - 2xy' + ()., - 1)y = O. (6.13)
Substituting (6.11) into (6.13) and collecting powers of x one obtains
(k + 2)(k + 1)ak + 2 - 2kak + (A - 1)ak = °
which leads to the recurrence formula
2k - ()., - 1)
(6.14)
ak + 2 = (k + 2)(k + 1) ak •
This determines the coefficients of the higher powers in (6.11) from those
of the lower powers. By comparing the ratio ak + z/a k for large values of k
with the ratios bk+ z/b k in a power series expansion of ex2 , we can show that
the function (6.11) with (6.14) will be divergent like ex2 , as x ~ ± 00. Thus,
the only way that we can construct solutions y(x) of the form (6.12) which
converge in the desired manner at ± 00 is to choose values for Asuch that the
series (6.11) terminates. This we can accomplish by letting A - I = 2n for
n = 0, 1,2, ... , and we thus see that t/I(x) can be an element of S only if
A = 2n + I or En = n + 1, n = 0, 1,2, .... (6.15)

The solutions of the differential equation (6.13) corresponding to the
eigenvalues (6.15) are also labelled by n and denoted by y(x) = Hn(x). Thus,
we obtain a family of equations
H~ - 2xH~ + 2nHn = 0, n = 0, 1,2, ... , (6.16)
which are called the Hermite differential equations. Their solutions Hn(x)
are called the Hermite polynomials. It is easy to check that the polynomials
defined by:
(6.17)
fulfill (6.16). Equation (6.17) is called the Rodrigues formula for the Hermite
polynomials.
Therewith we have found for the transition coefficient (6.7):
(6.18)
The factor N n will be chosen so that
(n/n) = f dx (n/x><x/n) = 1 ~n 12 f dx e- Hn(x)Hn(x) = 1. (6.19)

X2
The orthogonality of the eigenvectors of the self-adjoint operator H can

now be written in the form
N::Nn(n'/n) = f
dx e- X2 H n,(x)Hn(x) = ° for n # n', (6.20)
The second equality of (6.20), which we obtained as a consequence of the

orthogonality of the /n), is called the orthogonality relation for the Hermite
polynomials. The normalization factor N n has to be calculated using the
special properties of the Hermite polynomials defined by (6.17). This is done
using standard methods in the Appendix to this section. From (A.12) one
obtains
(6.21 )
N n and therewith the transition coefficients <x /n) in (6.16) are only deter-
mined up to a phase factor, i.e., a factor eia(n) of modulus one, which can still
vary with n. It is standard to choose these factors equal to ( + 1). Thus we

obtain
(6.22)
Appendix to Section 1.6
The normalization factor N n in (6.19) is calculated in the following way: We

consider the function f(t) = e-t2+2tx. Considered as a function of the
complex variable t, f(t) is an entire function and can be expanded into a
Taylor series of t:
co
e-t2+2tx= Lan(x)t n forall Itl<oo, (A.1)
"=0
where the expansion coefficients depend upon the parameter x. These
coefficients are given by
= ~ d"f(t) I (A.2)
an n.' d t n t=O
.
As f(t) is an analytic function for every value of the parameter x we can use
the Cauchy theorem which states that
d"f(t) = ~ 1
fez) dz
(A.3)
dt" 2ni :f(z - t)"+1 '
where the integration is along a contour enclosing the point z = t. Applying

this tof(z) = e-z2+2zx and using (A.2) we obtain at t = 0
_ 1 ~e- z2 + 2zx
a" - -2. +1 dz. (AA)
m z"
On the other hand, we can use (A.3) for the function fez) = e- z2 and
obtain
d" -x 2
dx" e
n.
= 2ni
'f (z _
e -z2
X)"+l
d
Z,
(A.5)
where the contour of integration encloses the point z = x, which can be in

particular any point on the real axis.
On the left-hand side of (A.5) we use (6.17). On the right-hand side we
perform a change of integration variable z = x - t, dz = -dt. Then (A.5)
goes over into
(A.6)
where the contour of integration encloses the point t = o. Dividing both

sides by e- x \ _1)" one obtains
n.
HnCx) = 2ni
, #e -t 2 + 2tx
t"+ 1 dt. (A.7)
Comparison of (A A) and (A.7) shows that the coefficients in (A.l) are given
by
and (A. 1) can be written

1
I
00
e-t2+2tx = I" HnCx)t" for It I < 00. (A.8)
"=0 n.
The functionf(t) = e-t2+2tx is called the generating function of the Hermite
polynomials. Many properties of orthogonal polynomials are easily derived
using their generating function.
We are now ready to calculate the normalization factor N". Using (A.8)
we obtain
f-
+OO
00
dx e-t2+2txe-s2+2sxe-x2 = II
".
00
m=0
t"sm
-'-I
n. m . -
f+oo
00
dx e- x2 HnCx)H m(x). (A.9)
The left-hand side of (A.9) can be written as
e 2st f_+: dx e-(x-(s+t))2 = e 2st f_+: d~ e-~2 = e 2st fi· (A. 10)
Expanding e 2st into a power series one obtains

2V(stY
left-hand side (A.9) = fi v=o
L
00
v! (A.ll)
Equating equal powers of sand t on the right-hand side of (A.9) and (A.ll)
one obtains
t+: dx e- x2 H"(x)H"(x) = fin! 2". (A. 12)
For n '" m one again obtains (6.20). Because of the property (6.20) and
(A.12) the HnCx) are called orthonormal on the interval - 00 < x < + CIJ
with respect to the weight function (fin! 2")-le- x2 .
Other important properties obtained from the generating function are the
recurrence relations. To obtain the recurrence relation for the Hermite
polynomials we differentiate (A.8) with respect to t:
Equating powers of t" one obtains:

(A.13)
This recurrence relation can be used to calculate the Hermite polynomials
step by step starting from Ho(x) = 1, H 1(x) = 2x.
In Chapter II we will also need some properties of the Fourier transforms
of the Hix) which are easily derived from the generating function. Multi-
plying (A.8) by eiYX-x2/2, where - 00 < y < + 00, and integrating we obtain:
J + 00 e2xt-t2+iyx-x2/2 dx = Jeiyx-x2/2 dx L n!1 H.(x)t"

-00 "
(A. 14)
The integral on the left-hand side is calculated using the substitution

IJ = x - 2t and gives (cf. Problem 11):
(A. 15)
For the right-hand side of this equality we use (A.8) with t replaced by (it).
Then we obtain from (A.14):
foe- Y2 /2 "~o ~Hn<y)

00 (it)"
=
00
"~on!
t" J+
-00
00
e,yx e -x 2/2H"(x)dx.
•
(A. 16)
Comparing coefficients of identical powers of t we obtain
(A.17)

In the previous sections of this chapter we presented the mathematics of
quantum mechanics as an algebraic structure. Generalized eigenvectors were
introduced by analogy to the basis vectors of a three-dimensional space and
it was asserted that they could be treated very much like ordinary vectors.
However, they are not vectors of the space but are in fact continuous
functionals on the space. In this section we define them and explain their
distinction from ordinary vectors.
Let be a complex linear space. A functional on is a mapping F from
the space into the complex numbers C, F: -+ C. (If is a real space then
15 This section can be omitted in first reading. It gives a mathematical explanation of notions
used in Sections 1.4 and 1.5 and is not essential for an understanding of the following chapters
except for Chapter XXI.
the mapping is into the real numbers IR.) It is thus the analog of a complex-
valued function of a real variable x E IR, F: IR ~ C, only now the variable is
not a real number but a vector </J E <1>. If F satisfies
F(a</J + f3t/J) = a*F(</J) + f3*F(t/J) for all </J, t/J E and a, 13 E C,
(7.l)
then F is called an antilinear functional. If F satisfies
F(a</J + f3t/J) = aF(</J) + f3F(t/J), (7.1a)
then F is called a linear functional. (If is a real space there is no distinction
between linear and anti linear mappings.) We will consider here antilinear
functionals rather than linear ones (note that in the mathematical literature
one usually considers linear functionals).
An example of an anti linear functional on a linear space is given by
F(</J) = (</J, t/J), (7.2)
where t/J is a fixed element t/J E and (</J, t/J) is the scalar product of t/J with
</J where </J varies over (cf. Problem 12). Because of this example and because
in the general case we want to consider a functional to be a generalization of
the scalar product, one uses for the antilinear functional F( </J) the symbol
F(</J) = <</JIF). (7.3)
Any two linear functionals, F 1 and F 2' on a linear space may be added
and multiplied by numbers according to
(aF 1 + f3F 2 )(</J) = aF 1 (</J) + f3F 2 (</J), a, 13 E C, (7.4')
or, using the notation (7.3),
(7.4)
The functional aF 1 + f3F 2 defined by (7.4) is again an antilinear functional
over (Problem 13). Thus the set of antilinear functionals on a linear space
 forms a linear space. 16 This space is called the conjugate space or dual
space (precisely, the algebraic dual or algebraic conjugate space) to the space
 and is denoted by x •
Let be finite-dimensional, dim = n, and let e;, i = 1,2, ... , n be a basis
for <1>; let F be an arbitrary antilinear functional on <1>. Let us denote by J;
the complex numbers
J; = F(e), i = 1,2, ... , n. (7.5)
The functional F at an arbitrary :3 </J = I7= 1 </Jiei can then be written as
n n
F( </J) = I F( </Jie) = I </J*i F( e) = I </J*iJ;. (7.6)
i= 1 i= 1
16 This is the analog to the statement that the set of linear operators on a linear space forms
an algebra (i.e., a linear space in which a multiplication is defined). (7.4) is the analog to the first
two equations of (3.4).
One has thus a one-to-one correspondence between the set of anti linear
functionals on a finite-dimensional space and the sequence of numbers
(fl' f2' ... ,fn)' With these numbers one can now define a vector f in by
n
f= I,!;e i · (7.7)
i; I
Then the right-hand side of (7.6) is the scalar product of the vectors tjJ and
f and we have shown that in a finite-dimensional linear scalar product space
there is a one-to-one correspondence between the functionals F and the
f E <1>:
(7.8a)
such that the value of the functional F( tjJ) at the element tjJ E is given by the
scalar prod uct :
F(tjJ) = (tjJ, f). (7.8b)
This shows that in the finite-dimensional case the symbol (7.3) can always
be identified with the scalar product 17
<tjJIF) = (tjJ, f). (7.8c)
When the identification (7.8) can be made, then is said to be self-duaP 7
and we write
(7.8d)
For infinite-dimensional scalar product spaces such an identification is in
general not possible. As we have already seen that the scalar product always
defines an antilinear functional, we can identify those antilinear functionals
Ff, whose values at any tjJ E can be written as the scalar product Ff(tjJ) =
(cp, f) with a fixedf E <1>, with the vectorsf: F f +--+ f. Then we have
(7.9)
For the understanding of functionals on infinite-dimensional spaces
topological notions are of great importance. So far we have only discussed
how algebraic structures are imposed upon a set. The linear scalar product
space defined in Section 1.2, which we want to call 'I' if it has no other struc-
ture, is a purely algebraic notion. One can now impose upon it a topological
structure. Topological structures are defined by giving a specific meaning to
the notion of convergence of infinite sequences. IS
The Hilbert space convergence of a sequence of vectors CPl' tjJ2' tjJ3"'"
tjJv' ... to the vector tjJ E ~ denoted by
tjJv ~ tjJ for v -> 00, (7.10')
17 This identification is not necessary and often it is useful not to identify the functionals
with elements of <1>.
IS There are topological spaces for which the definition of convergence of sequences is
not sufficient to define the topology. But for the spaces which we shall consider here (with
first axiom of countability) this definition is sufficient.
is defined as
IlcP, - cP I --+ 0 for v --+ 00. (7.10)
The -spaceconvergence is defined with the help ofthe algebra of operators
and is therefore dependent upon it. For the algebra generated by P, Q, H
of (6.3) and (6.4) it is defined by19
cP, --+ cP for v --+ 00 iff (7.11')
IIHP( cP, - cP) II --+ cP for v --+ 00 for every p = 0, 1,2,3,.... (7.11)
As (7.11) includes (for p = 0) the property (7.10) it is clear that from
cP, --+ cP follows
.te cP
cP, --+ (7.12)
but not vice versa. The convergence (a topology) defined by (7.11) is called
stronger (finer) than the convergence (topology) defined by (7.10), and the
convergence defined by (7.10) is called weaker (coarser) than the convergence
defined by (7.11). Yt' is the space which contains in addition to the elements
of \fI all limit elements of ~-convergent sequences. is the space which
contains in addition to the elements of \fI all limit elements of ~-convergent
sequences. As because of (7.12) every -convergent sequence is also Yt'-
convergent but not vice versa, we have
 c £.' (7.13)
The antilinear functionals that we will consider in the case of infinite-
dimensional spaces will always be continuous functionals. An antilinear
functional is continuous iff from
<C
cP, --+ cP follows F(cP,) --+ F(cP) for v --+ 00, (7.14)
where ~ denotes convergence for complex numbers. 20 x, Yt' x, etc. will
always denote spaces of continuous functionals.
We can now consider the set of continuous linear functionals on Yt' and
on <1>.
 x is the set of all F with the property that
F( cP,) ~ F( cP) for all cP, ~ cP.
Yt' x is the set of all F.te with the property that
F.te(cP,) ~ F.te(cP) for all cP, ~ cP.
The condition fulfilled by F E Yt' x is more stringent than the condition
fulfilled by F E x because according to (7.12) there are more sequences for
which cP, ~ cP than sequences for which cP, ~ cP. Therefore
Yt'X c x • (7.15)
19 This is the weakest topology or convergence that makes all elements of the algebra con·
tinuous operators.
20 A sequence CI> c 2 , ••. of C v E C converges to a number C iff Ic, - cl--+ 0 for v --+ ro.
As mentioned above, infinite-dimensional linear spaces do not in general

have the property (7.8d). However, the infinite-dimensional Hilbert space
has the following remarkable property expressed by the Frechet-Riesz
theorem: For every JIt'-continuous functional F£' there exists an f E Jlt'such
that
(7.16)
Thus for JIt' one can make the identification (7.8) as for finite-dimensional
spaces and obtains from (7.13) and (7.15) the triplet of spaces:
(7.17)
<
called a Gelfand triplet or rigged Hilbert space. The symbol ¢ IF) is therefore
an extension of the scalar product to those F E x which are not in JIt'.
One can now consider antilinear continuous functionals qJ on x and
denote the space of all qJ by x x. For a large class of linear topological
spaces (called reflexive) there is a natural one-to-one correspondence
between a ¢ E and qJ E x x given by
qJ(F) = F(¢) = <¢IF). (7.18)
<
As ¢ IF) is an extension of (f, ¢) = (¢, f) we can consider the functional
qJ at the vector F E x as an extension of the scalar product (f, ¢). It is
therefore convenient to write qJ(F) == <F I¢), where FE x is now the
variable, i.e., <F I¢) is the value of the functional qJ on the space x at the
vector F. With this definition of <FI¢), (7.lS) reads
<FI¢) = <¢IF) (7.18')
giving an extension of the property (2.2b) of the scalar product.

In order to define the generalized eigenvectors which appear in the spectral
resolution (4.4c) we will have to consider the linear operators on <1>. We will
only consider operators which are continuous on <1>. In analogy to the defini-
tion of a continuous functional we say that an operator A: -+ is con-
tinuous iff for all convergent sequences
~ ~
¢v -+ ¢ follows that A¢v -+ A¢.
The notion of continuity thus depends-like all topological notions-on the

definition of convergence. An operator which is a continuous operator with
respect to the -convergence need not be a continuous operator with respect
to the JIt'-convergence. (Many physical observables cannot be represented
by continuous operators on the Hilbert space JIt' but they can still be con-
tinuous operators on <1>.)
For every continuous linear operator A on one can define an adjoint
operator A x on x by
(7.19)
One can prove that A x is a linear continuous operator on x. This is the
extension of the Hilbert space-adjoint operator defined by (cf. Section 1.3):
(ep, AY) = (Aep, f) for all ep E (7.20)
(which in general is not defined for allf E £).
If one has two operators A and A with A defined on a subspace of the space
on which A is defined and with the property Aep = Aep for all ep E subspace,
then one writes A c A and calls A an extension of the operator A. For an
operator A on which is self-adjoint and has a unique self-adjoint 21 exten-
sion A to .Yt' one has therefore the following triplet of operators:
A c A = At c Ax (7.21)
on the triplet of spaces
A functional F over is called a generalized eigenvector of the operator A

on with eigenvalue w iff
<AepIF) = w<epIF) for all ep E <1>. (7.22)
The generalized eigenvector F is also denoted
IF) = Iw). (7.23)
Another precise form of writing (7.22) is
AX Iw) = wlw), (7.24)
which is often just written as
Alw) = wlw). (7.25)
If F ~ f E .Yt' then (7.22) becomes
(Aep, f) = (ep, Atf) = weep, f) for all ep E <1>, (7.26)
which for A self-adjoint (as is dense in £) is identical with the definition
of an eigenvector (cf. (3.6b»:
Af = wf. (7.27)
The definition (7.22) of a generalized eigenvector and its generalized eigen-
value is thus a generalization of the definition (7.27) of a proper eigenvector
and its proper eigenvalue. This generalization need not have all the properties
of ordinary eigenvectors and eigenvalues. In particular, w in (7.22) need not
be real even if A is (essentially) self-adjoint.
The eigenvectors Ix) that appear in the continuous eigenvector expansion
are generalized eigenvectors in the sense of (7.22). The coordinate <x Iep) =
ep(x) of the vector ep E along the basis vector Ix), or the value of the function
21 An operator which is defined on $, is Hermitian on $ as defined in Section 1.3 and has a

unique self-adjoint extension to .Yf is calJed essentialJy self-adjoint.
Problems 39
¢ at the number x, is thus the complex conjugate of the value of the functional
Ix) E x at the vector ¢ E <1>: (xl¢) = (¢Ix)*. Using (7.18), one can also
say that (x I ¢) is the value of the functional ¢ E x x = at the vector
(XIEx.
1.8 How the Mathematical Quantities Will Be Used
We conclude this chapter by a table which gives the correspondences between

the mathematical objects introduced in this chapter and the physical quanti-
ties which they will represent. To establish, explain, and utilize this corre-
spondence will be the subject of the following chapters of this book.
Mathematical Image Physical Quantity
vector ¢ E <ll (modulo a phase factor) pure physical state

linear operator A on <ll physical observable
eigenvalues (spectrum) of A values obtained in a measurement of the
observable A
eigenvector IA) of A with eigenvalue A state in which measurement of A gives the
value A, eigenstate, bound state
generalized eigenvector IA) of A with scattering state
real eigenvalue A
generalized eigenvector Iw) of A with resonance state with resonance energy E and
complex eigenvalue w = E - ir/2 width r
scalar product (A I¢) with IA) E <ll, probability amplitude
<PE<ll
modulus squared of it I(AI¢)1 2 probability to obtain the value Ain a
measurement of the observable A on a
physical system in the state ¢
functional <A I¢) wave function
modulus squared of it I<AI¢W probability density for measuring the value A
for the observable A in the state ¢
diagonal matrix element or average value for the measurement of A in
expectation value (¢, A¢) in the state ¢
matrix element (1/1, A¢) transition amplitude
its modulus square 1(1/1, A¢W transition probability for transition caused
by A from state ¢ to state 1/1
Problems
1. Let C" denote the set of all sequences x = (~1' ~2' ... , ~n) with ~i E C (complex
numbers). Define addition and multiplication with elements of C by the formulas
(~l' ~2"'" ~n) + ('11' '12'"'" /]n) = (~1 + /]1' ~2 + /]2"'" ~n + /]n)'
a(~l' ~2"'" ~n) = (a~1' a~2"'" a~n)'
(a) Show that en is a linear space.

(b) Let x = (~1' ... ' ~n)' Y = ('11, ... , '1.) be two elements of en. Show that (x, Y)
defined by
(x, Y) = I ~i'1i
i= 1
fulfills all conditions of a scalar product.

2. Let be en of Problem 1 and let a map A be defined by
where
~'i = I aik~k'
k=1
where a ik (i, k = 1,2, ... , n) are fixed numbers.
is called the matrix of the map A. Define another map B in the same way. Show that
these maps A and B are linear operators on the space en (i.e., fulfill the relations for
linear operators). Show that the set of operators form an algebra if multiplication and
addition of their matrices and multiplication of their matrices by a number are defined
by the usual rules for matrix multiplication.
3. Show that af (a E e, f E <1» is an eigenvector of a linear operator A if f is an eigen-
vector of A with eigenvalue..t What is the eigenvalue of the vector af?
4. Show that a Hermitian operator A has the following properties:
(a) All eigenvalues are real.
(b) Two eigenvectors 4>1 and 4>2 of A are orthogonal to each other if the eigenvalues
corresponding to them are different from each other.
5. Show that two vectors 4> and I/J are equal iff all their components with respect to a
basis system are equal.
6. Show that the Cauchy-Schwarz-Bunyakovski inequality, Equation (2.7), follows
from the definition (2.5) of a positive Hermitian form.
7. Let Q be the operator with continuous spectrum {x 1- oc < x < + oo} and let
1x) denote its generalized eigenvector. Define another operator P by:
I d
<xIP11/I)=-:-d <xll/J) forall I/JE (i2 = -1).
I x
(a) Show that P is a linear operator.
(b) Show that in order to have the operators Q and P well defined, i.e., 11P4>1I and
IIQ4>1I be finite for all vectors 4> E <1>, the components of every vector <xl4»
must be infinitely differentiable continuous functions which together with their
derivatives vanish at infinity more rapidly than any polynomial of x (i.e., <x 14»
must be functions of the Schwartz space S).
Problems 41
(c) Show that the operators P and Q fulfill the commutation relation
I
PQ - QP = ~ 1.
I
8. Let be a linear scalar product space with elements ¢, t/J, x, .... Let Ii), IEn),
i = 1,2, ... , En = E 1 , E 2 , ••• , be two discrete basis systems in and Ix), - CJJ < X < + CJJ
a continuous basis system of <1>. Show that the spaces of components Cin with elements
<il¢), <ilt/J), <ilx),··· and Cit) with elements (Enl¢), (En It/J), (Enlx),··· and S(x) with
elements <x I¢), <x I t/J), <x I X), ... are all isomorphic to the space and isomorphic to
each other.
9. Show that the generalized function <xlp) = (l/j2n)e iXP is a generalized eigen-
function of the operator defined in Problem 7:
I d
<xIPlp) = ~-d <xlp) = p<xlp)
I x
if P is real, but that it is not a well-behaved function, i.e., that Ip) ¢ <1>.
10. Show that the polynomials defined by
n = 0, 1,2, ...
satisfy the differential equation
H~ - 2~H~ + 2nH n = o.
11. Calculate the Fourier transform
¢(x) =;::;;:
1 f+oo dp eixp(fJ(p)
y' 211: - 00
of the function (fJ(p) = e- ap2 E S (a> 0).

12. Show that the scalar product (¢, t/J) with t/J E fixed and ¢ E variable defines
an antilinear functional F.,(¢) == (¢, t/J) on the space <1>.
13. Let F 1 and F 2 be linear functions over <l>, IX, f3 be complex numbers, show that
(IXF 1 + f3F 2) defined by
for all ¢ E is again a linear functional over <1>.

14. Let p 2 be the operator defined by Equation (5.25) and P the operator defined by
I d
<xIPI¢) = ~- <xl¢)
I dx
with the same requirements for the functions <x I¢). Show that as a consequence of
(5.26) P and p 2 are self-adjoint operators. Are P and p 2 self-adjoint operators if (5.26)
is not fulfilled? Hint: Use integration by parts.
15. In Section 1.7 it was shown that
Jf' c x by proving Jf' x c x
and using the Frechet-Riesz theorem .Yf = .Yf x. Give an alternate proof of .Yf c <l> x
without using .Yf = .Yf x by proving the following:
(a) Show that from
.tf
hn -> h follows (h n , f) -> (h, f)
for allf E Yt'. Hint: Use the Cauchy-Schwarz-Bunyakowski inequality.
(b) Show that every f E.Yf defines an element of FE <l>x by F(<p) = (<p, f) for
every . E.g., show that F, which is anti linear by Problem 12, is continuous.
Hint: Show that from gn ~ 9 follows F(gn) -> F(g) using the result of part (a).
CHAPTER II
Foundations of Quantum
Mechanics-The Harmonic
Oscillator
This chapter, the longest in the book, introduces three of the basic assump-
tions of quantum mechanics and then illustrates them, using mainly the
example of the harmonic oscillator. Though some historical remarks are
included, neither the historical development nor any other heuristic way
towards quantum mechanics is followed. The basic assumptions are
formulated, explained, and applied. In Sections 11.2, 11.4, the basic assump-
tions are introduced; in Sections 11.3, 11.5, 11.7 the harmonic oscillator
is used to illustrate them. Section 11.6 contains the derivations of some general
consequences and might be omitted in first reading. The discussion for the
continuous spectra, important for the description of the scattering and
decay phenomena in the second part of the book, is given in Section 11.8.
Several remarks throughout this chapter emphasize the particular problems
connected with generalized eigenvalues and eigenvectors and our unified
treatment of continuous and discrete spectra. In Section 11.9 we are ready
to explain the physical meaning of the quantum-mechanical constant of
nature, h.
ILl Introduction
Physicists believe that there is something in nature, or in each restricted

domain of it, that may be "understood"; that there is a structure in nature.
To "understand" means to bring this structure into congruence with some
structure in our mind, with a structure of thought objects, with a structure
43
44 II Foundations of Quantum Mechanics-The Harmonic Oscillator
that has been created by our minds. For physics this structure of thought
objects is a mathematical structure. So to understand part of physical nature
means to map its structure on a mathematical structure. To obtain a physical
theory, then, means to obtain a mathematical image of a physical system, by
which we mean any suitably restricted domain of physical nature (e.g., just
an atom, or even just a set of substates of it, or all atoms and molecules).
For the domains of quantum physics the mathematical structures are
algebras of linear operators in linear spaces. The discovery of this, the
fundamental properties of the algebra, and the other basic assumptions of
quantum mechanics was a very difficult process, the history of which we do
not want to describe here. It is much easier to start with the basic assumptions
and to show that conclusions derived from them really describe some parts
of nature. This is the path we want to follow here.
11.2 The First Postulate of Quantum Mechanics
The first basic assumption or postulate (" axiom") of Quantum Mechanics

IS:
I. A physical observable, defined by the prescription for its measurement, is

represented by a linear operator! in a linear space. The mathematical image
of a physical system is an operator *-algebra in a linear scalar-product space
JIf. The algebra is generated by some basic physical quantities, and the
multiplication is defined by algebraic relations.
The content of this postulate will only become clear after one has seen it
in operation in a number of concrete physical examples. This postulate is
the result of a long and tedious process of abstraction and one of the most
glorious achievements of science.
The elements of the algebra, the linear operators representing a physical
observable, are not to be considered as anything in particular-like differen-
tial operators acting on well-behaved functions. Rather, the algebraic symbol
stands for something which is defined only in its relationship to other such
symbols through mathematical operations in which these symbols take
part. A single algebraic symbol is thus empty and its properties are assigned
to it by its relation to all other symbols. And the more (independent) relations
one imposes, the more specific becomes the property of the mathematical
entity. The meaning of a single algebraic symbol is determined by the way in
which the algebraic language as a whole is used.
To illustrate axiom I we use one of the simplest examples, the one-dimen-
sional harmonic oscillator. The one-dimensional harmonic oscillator can
1 In Chapter XIX we shall also have to admit semi linear operators.

11.2 The First Postulate of Quantum Mechanics 45
be defined as the physical system whose mathematical image is the algebra

of operators that is generated by the following Hermitian operators:
H, representing the observable energy.
P, representing the observable momentum.
Q, representing the observable position.
The defining algebraic relations that are fulfilled by these operators are
Ii
[P, QJ == PQ - QP = --;- I, (2.1 a)
I
(2.1 b)
where Ii, ji., and OJ are numbers, I is the unit operator, and i = j=1. The
physical meaning of the numbers ji. and OJ will be found by correspondence
to the classical system to be the mass and angular frequency of the oscillator,
which are system constants. The physical meaning of the universal constant 2
Ii will be discussed in Section 11.9.
The justification for the above statements is that there exist in nature
physical systems on which one can perform measurements that lead to
numbers that can be calculated from the above assumptions. Therefore
it is reasonable to say that these physical systems, which are as always
"idealized physical systems," have as their mathematical image the above
mathematical structure. The name "harmonic oscillator" for this physical
system arises from the correspondence to a classical system that has the
same name.
The classical-mechanical system called a one-dimensional harmonic
oscillator has the following structure (Figure 2.1). Two mass points, with
masses m 1 and m 2 , are acted upon by a force F along the line joining the two
points. The force F is proportional to the difference between the distance r
and its equilibrium length ro:
F = -k(r - ro) = -kx, (2.2)
where x = r - ro' The potential energy for this system is then
V = +tkx2, (2.3)
and the kinetic energy is
1 2
Ekin = 2ji. P , (2.4)
2 2rrh = h is called Planck's constant.
m, m2
0>--------0
1 1
I. _I
Figure 2.1 Model of a one-dimensional oscillator.
46 II Foundations of Quantum Mechanics- The Harmonic Oscillator
Figure 2.2 Classical one-dimensional oscillator.
where we have introduced the reduced mass
(2.5)
and the momentum

dx
p = /1 dt' (2.6)
We can therefore define a classical harmonic oscillator as a system whose

potential energy is proportional to the square of the distance from its
equilibrium position, or as a system whose total energy is
1
E = 2/1 p2 + tkx2. (2.7)
As is well known, such a physical system is an idealization of the mechanical

system in nature that consists of two masses connected by a spring (Figure
2.2). Such a mechanical system performs oscillations with an angular
frequency
(2.8)
If we write Equation (2.7) in terms of w by solving Equation (2.8) for k, we

have
1 /1W2
E = ~ p2 + __ x 2. (2.9)
2/1 2
This expression looks exactly like one of the relations defining the quantum-
mechanical harmonic oscillator. The important difference is that in the
classical case the energy E, the momentum p, and the position x are real
numbers, whereas in the quantum-mechanical case these quantities are
represented by the operators H, P, and Q, respectively.
For real numbers a and b one always has the relation ab - ba = 0; thus
px - xp = O. (2.10)
Generally, for operators A and B, AB - BA i= 0, i.e., operators do not, in
general, commute. For the particular case of the momentum and position
operators one has the relation (2.la).
Fundamental relations like (2.la) and (2.1b) cannot be derived; they can
only be conjectured. This is true not only for quantum mechanics but also
IL2 The First Postulate of Quantum Mechanics 47
for classical physics. Thus the relation (2.9) is essentially the result of con-
jectures that go back to Newton.
The process of conjecture in physical theory can be described simply as
follows: First one considers the experimental data in a certain domain of
physics, e.g., the atomic domain. Then one tries to construct a mathematical
structure from which one can calculate numbers agreeing with the experi-
mental data. The algebra of operators defined by the relations (2.1) is such a
mathematical structure for the quantum-mechanical harmonic oscillator.
The conjecture (2.1b) was simple for us, since it was obtained by correspon-
dence with the classical case. The physical meaning of the system constants
J1 and w follows from the classical analogue. The conjecture (2.la), however,
was not so simple. The replacement of the classical relation (2.10) by the
quantum-mechanical relation (2.la), and thus the realization that the
quantum-mechanical observables are not represented by real numbers but
by operators, was one of the greatest achievements of science. The relation
(2.la) is called the Heisenberg (or canonical) commutation relation. The
physical meaning of this relation will become clear when we study its conse-
quences in the following sections.
Since we do not want to develop quantum mechanics along its

historical path, we shall give only a very brief and partial sketch of
the motivation for (2.1a). It was known from experiment that light
is emitted from atoms at discrete frequencies. At the time the
developments leading to (2.1a) took place, the atom was pictured
as consisting of a positively charged nucleus with a very small radius
and negatively charged electrons revolving around it at a distance
of about 10 - 8 cm. Describing this picture in terms of classical
physics lead to a dilemma. A revolving and therefore accelerating
electron would emit radiation, leading to a continuous loss of its
energy and a continuous change in the frequency of the emitted
radiation until the electron would finally, and very rapidly, fall
into the nucleus. This contradicted the experimental situation that
atoms are quite stable, and the frequency of the emitted radiation
can only take certain definite values (spectral lines), which are
characteristic of the atom. Hence classical theory could not describe
the observations in the atomic domain, if the above picture of the
atom was kept. One did not wish to abandon the picture of a
miniature planetary system for the atoms, since replacing the
Hamiltonian by another classical Hamiltonian that would take only
discrete values would be virtually impossible. To obtain only dis-
crete values for the Hamiltonian, some extra conditions were im-
posed on the classical theory that were not only supplementary to,
but also in contradiction with the classical theory. This was the
so-called "old quantum theory" (Bohr).
Quantum mechanics originated in the idea of Werner Heisenberg
and Max Born (1925) and Erwin Schrodinger (1926) to interpret the
Hamiltonian as a mathematical quantity that would take only

discrete values. From an analysis of the harmonic oscillator,
Heisenberg and Born and Jordan concluded that in order for the
energy to be a system of discrete values, the momentum and position
should be given by the matrices 3
0 ji 0 0
-ji 0 J2 0
p = -iJJ1~h 0 -J2 0 J3 , (2.11)
0 0 -J3 0
0 ji 0 0
ji J2
Q-f£
0 0
2J1w
0 J2 0 J3 (2.12)
0 0 J3 0
where 2nh = h is Planck's constant. Schrodinger built on a partial

theory due to de Broglie, according to which free particles should
diffract like light and should therefore be described by a wave
function that satisfies a wave equation. From this he concluded that
the momentum should be the differential operator
p=~ ~ (2.13)
i ax
applied to the wave function. Both the Heisenberg and the
Schrodinger Ansatz lead to the relation (2.1a); and soon after the two
formulations were made known, it was shown that they were two
different realizations ofthe same mathematical entities, namely linear
operators in a linear space. We shall give a brief description of
subsequent historical developments in Section 11.5. 4
The quantum-mechanical harmonic oscillator is an idealization of micro-

physical systems in nature; e.g., diatomic molecules are, under certain
conditions, quantum-mechanical harmonic oscillators. The simplest possible
geometrical (classical) picture of the diatomic molecule is two atoms, of
masses m 1 and m 2 , bound together by an elastic force.
There is an essential difference between the descriptions, in everyday
language, of the physical systems whose idealizations are the classical
3 This particular representation of the P and Q as matrices of operators with respect to a

certain basis of vectors in an inner-product space will be derived in the next section (Section 11.3).
4 A few historical remarks will appear throughout the book. For a description of the historical
developments see, e.g., Jammer (1966).
II.2 The First Postulate of Quantum Mechanics 49
harmonic oscillator and the quantum-mechanical harmonic oscillator.

Everyday language gives quite an adequate description of classical systems,
e.g., two carts with masses connected by a spring. Everyday language is,
however, insufficient to describe quantum-mechanical systems. The quantum-
mechanical harmonic oscillator is best described by the mathematical
structure given by (2.la, b). This is the justification for the above definition
and the conjectures that led to it.
The procedure of conjecturing the mathematical relations outlined above
for the harmonic oscillator works for many microphysical systems, so we
summarize it again:
1. Obtain a classical picture of the microphysical system, and express
the observables as functions of position and momentum coordinates.
2. Replace the real numbers corresponding to momentum p and position
x in the expressions for other classical observables by the Hermitian
operators P and Q, respectively. The operators P and Q satisfy the
Heisenberg commutation relation, and from them we obtain the
algebra of observables for a quantum-mechanical system.
Not all classical observables can be expressed in terms of momentum and

position: the intrinsic angular momentum or spin of a top, for example,
cannot be expressed as a function of the momentum and position of the top.
For these observables the above procedure of correspondence between the
classical and the quantum-mechanical observables has to be generalized.
We can give a vague formulation of this conjecturing procedure: in the
transition from the mathematical description of the classical system to the
mathematical description of the corresponding quantum system, we replace
the numbers representing classical observables by operators that represent
the corresponding quantum observables. The mathematical properties of
these operators are specified by the mathematical (in particular algebraic)
relations between them. These relations between the observables are obtained
from the relations between the classical observables and some additional
relations that specify the property of noncommutativity. These relations
can often be obtained from analogy to known relations between similar
quantities.
There exist (in elementary-particle physics for example) many quantum-
mechanical observables without known classical counterparts. The defining
relations between these observables can then be obtained only phenomeno-
logically from a comparison between the consequences of these relations
and the experimental data.
We shall discuss many examples of this conjecturing procedure. Rarely is
the choice of the relations unambiguous, and the ultimate justification of
the defining relations, like that of any fundamental theoretical assumption,
is their success in providing agreement between theoretical consequences
and experimental data.
The representation of physical observables by operators poses the problem
of relating these mathematical quantities to experimental data, which are
numbers. The solution of this problem is given by a number of further basic

assumptions or axioms, which provide the actual interpretation of quantum
mechanics. We shall introduce these axioms at suitable places when they
are needed.
11.3 Algebra of the Harmonic Oscillator
We now give a mathematical description of the quantum-mechanical

harmonic oscillator. To find the mathematical properties ofthe algebra of the
harmonic oscillator means to find how the operators of this algebra act on
the vectors of the space Yt'. In order to do this we introduce a new operator
a and the operator at (which is its adjoint):
(3.1)
Then we introduce the new operator

N = ata, (3.2)
which is Hermitian. Inserting (3.1) into (3.2) we get
N = pw Q2 + _1_ p2 _ ~ (PQ _ QP). (3.3)

2h 2pwh 2h
Using the relations (2.1) we obtain
1 1
N = wh H - :2 1. (3.4)
Similarly we calculate
I 1
aat = -H + -1. (3.5)
wh 2
Therefore as a consequence of the canonical commutation relation (2.1a)
we obtain for a and at the commutation relation
(3.6)
Furthermore, we can express the energy operator H in terms of a and at to
obtain
(3.7)
Now we want to find the eigenvalues and eigenvectors of Hand N. As

the eigenvectors of N are eigenvectors of H, it is sufficient to find all the
eigenvectors of N.
II.3 Algebra of the Harmonic Oscillator 51
Let us assume that there is at least one eigenvector of N in Yt', and call its
eigenvalue A.. (This is not a trivial assumption, since there are many operators
that do not have eigenvectors in Yt'. In fact it is an additional assumption
without which Equation (2.1a) is ambiguous.) Let CPA be this eigenvector;
thus
N cP A = A.cp A • (3.8)
Then calculate N acp A using Equation (3.6),
NaCPA = (ata)acpA = (aa t - I)acpA = a(ata - J)CPA

= a(N - I)CPA = a(A. - l)cpA = (A. - l)aCPA'
Thus
(3.9)
so that either acp A = 0 or acp A is an eigenvector of N with eigenvalue A. - 1.

Suppose the latter is the case. Then we can define
(3.1 0)
Repeating the above calculation on CPA _ l' we find that

(3.11 )
is either zero or an eigenvector with eigenvalue A. - 2. We continue in this

way and obtain a sequence of vectors
(m = 0, 1, 2, ... ), (3.12)
which are eigenvectors of N with eigenvalue (A. - m), as long as CPA-m =f. O.
In a similar way one calculates N(atcpA) to get
(3.13)
Hence cP HI == at cP A is either the zero vector or an eigenvector of N with

eigenvalue A. + 1. It is easily shown that always cP HI =f. O. Suppose at cP A = O.
That means
(3.14)
But
(3.15)
and by definition,
(at CPA' atcp;) = (CPA' aatcpA)' (3.16)
By the commutation relation for a and at we have
( cP A' aa t cP A) = (cp A' at acp A) + (cp A' cP A)
= IiaCPAI1 2 + IIcp AI1 2 > 0, (3.17)
since qJ;. =I O. Thus qJ H 1 is always an eigenvector of N with eigenvalue

A + 1. Again by repeating the above process one obtains a sequence of
vectors qJ).+n (n = 0, 1,2, ... ), which are eigenvectors of N with eigenvalue
A + n. These are always nonzero, as follows from the above proof.
We now determine when qJ;'-m can be zero. We calculate
and
Since the norm is always nonnegative, it follows that
A- m = IlaqJ;.-mll: > O. (3.18)

IlqJA-mll -
Therefore the sequence of eigenvectors qJ A _ m must terminate after a finite
number of steps, and there must exist one vector qJo such that
aqJo = O.
The vector qJo is an eigenvector of N with eigenvalue zero, since N qJo =
ataqJo = atO = O. Now we define the normalized vectors
(3.19)
From the considerations above we know that the vectors ¢n are eigenvectors
of N, with eigenvalue n, and the en are chosen such that the ¢n are normalized.
Thus we have
N¢n = n¢n' II¢nll = 1. (3.20)
The Cn are calculated as follows: From Equations (3.20) and (3.19),
1 = II¢n11 2 = (atn¢o, atn ¢o)ICnI2 ,

but
so
1- ICn 12 (A. t A. )
-ICn-l 12 'f'n-l,aa'f'n-l'
II.3 Algebra of the Harmonic Oscillator 53
Using Equation (3.6),
1 = I~Cn1212 (¢n-l' (ata + l)¢n-l)

n-l
I Cn l2
= I Cn_ 1 12 (¢n-l' (N + l)¢n-l)
ICn l2
= n IC 12 (¢n-l' ¢n-l)
n-l
- n ICn 12 •
- ICn_ 1 1 2 '
hence Cn must be chosen so that

nlCnl2 = ICn_ 1 12. (3.21)
Since ¢o is normalized, Co = 1, and one solution of Equation (3.21) is then
Cn = /1,.
-J;! (3.22)
There are other solutions of (3.21) which differ from (3.22) by a factor of
modulus 1; we choose (3.22).
We may summarize the procedure as follows: Start with a normalized
vector ¢o E YE that has the following property:
a¢o = 0. (3.23)
Then apply to ¢o the operator at to obtain a system of eigenvectors of N,
¢n = ~ atn¢o, (3.24)
v' n!
with eigenvalues n = 0, 1, 2, .... These eigenvectors are normalized by the
construction above, and they are orthogonal to each other since they are
eigenvectors of a Hermitian operator with different eigenvalues (cf.
Problem 1.4). Thus we have the orthonormality relations
(¢n' ¢n,) = Dnn " (3.25)
and the ¢n form an orthonormal system in YE. The operators a and at are
defined on this orthonormal system by
a¢n = vIn¢n-l'
at", = ifl+1", (3.26)
'+'n v''' T ~'+'n+l'
Because of this property, a is called the annihilation operator and at the
creation operator. Both are often referred to as ladder operators. All the
elements of the algebra of the harmonic oscillator that are functions of a
and at are defined on the elements of this orthonormal system. Consider the
set of all vectors
(3.27)
n
where the cxn are complex numbers and the sum runs through an arbitrarily
large but finite number of n. This set of vectors forms a linear space called the
space spanned by the vectors ¢n' We will denote this space by Yf or when
we ignore details.
[More precisely, the space of all the I/J = I:; 0 cxn ¢n for which
I:; 0 Icx n l2 < c:fJ will be denoted by Yf (for Hilbert space), and the
space of all I/J for which I:; 0 I CXn 12(n + I)P < c:fJ for all p = 0, 1,
2, 3, ... will be denoted by <1>. Obviously
(3.28)
The advantage of over Yf is that all operators representing
observables can be defined on the whole space <1>, but not on all of
Yf. Though both spaces and Yf are mathematical generalizations,
<l> is more convenient for physics.]
As all observables are functions of a and at, and a and at are known by (3.26)
and (3.27) for all I/J, all observables for the quantum-mechanical harmonic
oscillator are in principle known on <1>. So the task of determining the mathe-
matical properties of the algebra of the harmonic oscillator is in principle
completed.
Let us calculate the diagonal matrix element of the energy operator H
between the vector ¢n for a fixed value of n. Using Equations (3.7) and (3.20),
we have
(3.29)
and
(¢n' H¢n) = hW(¢n' N¢n) + hW(¢n' 1¢n)
= hw(n + 1)I¢nI 2 = hw(n + 1)'
We call this number En:
En = hw(n + 1)' (3.30)
En are the eigenvalues of the operator H. The set of all eigenvalues is called
the spectrum of an operator. The spectrum {En In = 1,2, ... } of the operator
H is the energy spectrum of the harmonic oscillator.
11.4 The Relation Between Experimental Data and Quantum-

MechanicalObservables
We now wish to formulate the basic axiom of quantum mechanics that gives
the connection between the operator representing an observable and the
experimentally measured values for this observable. Before doing this, we
observe some fundamental differences in the experimental properties of the
classical harmonic oscillator and the quantum-mechanical oscillator.
Consider the classical picture in Figure 2.1. Assume that the system has no
energy-i.e., does not perform any vibrations, or vibrates with zero amplitude.
IIA Experimental Data and Quantum-Mechanical Observables 55
The system can be excited into vibration if, for example, one of the masses is
struck by another massive projectile. The oscillator can perform vibrations
of any amplitude, depending on the amount of energy transferred to it,
which in turn depends on the momentum of the projectile. Thus for the
classical oscillator one can prepare a projectile that excites the oscillator to
vibrate with any desired energy E. The energy is given by
E = J;ka 2 , (4.1)
where a is the amplitude of the vibrations
x(t) = a sin(wt + eo).
In the realization of the classical harmonic oscillator as two massive carts
(see Figure 2.2), the system may be excited, for example, by hitting one of
the carts with a third cart of suitable momentum.
The situation is quite different for the quantum-mechanical harmonic
oscillator. To see this we will consider an analogous experiment with micro-
physical systems. As is typical for experiments with microphysical systems,
one does not have a single physical system but an ensemble of systems, i.e., a
collection of systems that are in a certain sense identical. The ensemble of
microphysical harmonic oscillators that is used in this experiment is a gas of
CO molecules, and an electron beam serves as an ensemble of structureless
projectiles.
Remark: It should be remembered that the statement that a physical

system is structureless has no universal meaning. Any system can be
considered structureless under certain conditions but not under
others. For instance, in the domain of physics for which the kinetic
theory of gases is valid, the molecules can be considered as structure-
less systems, whereas when higher energies are considered, they are
seen to have structure. The electron is a system that can be con-
sidered to be without" energy structure" for all currently available
energies.
Experiments of this type are called energy-loss experiments, and are based on
the original experiment performed by Franck and Hertz in 1914.
Figure 4.1 shows a schematic diagram of such experiments. A beam of
electrons leaves a monochromator 5 with a very narrow energy band of
5 A monochromator is an apparatus that prepares a beam with very well-defined momentum

and, therefore, with very well-defined kinetic energy. An analyzer is the same kind of apparatus
used for a different purpose.
-
MONO
CHROMATOR r-!.. CO
GAS ~ ANALYZER
E., r-- DETECTOR
E.
Figure 4.1 Schematic diagram of an energy-loss experiment.

Daubl. tllttroslltlc '1IIlynl'

In dlgrtftl
(a)
,3
co
o I Z
S... ~p Vol!a~ . vol!
Energv loss (eV I
(b)
Figure 4.2 (a) Schematic diagram of double electrostatic analyzer. Electrons are
emitted from the thoria-coated iridium filament. They pass between the cylindrical
grids at an energy about 2.05 eV, and are accelerated into the collision chamber, where
they are crossed with a molecular beam. Those electrons scattered into the acceptance
angle of the second electrostatic analyzer pass between the cylindrical grids, again at
an energy from 0 to ;::::; 2 eV. The electrons pass the exit slit into the second chamber and
impinge onto an electron multiplier. (b) Energy spectrum of scattered electrons in CO
at an incident electron energy of 2.05 eV.
II.4 Experimental Data and Quantum-Mechanical Observables 57
energy Ee, and enters a collision chamber filled with an ensemble of CO

molecules (or any other physical system whose structure is to be investigated)
kept at a low temperature. Some of the electrons scatter into an analyzer
which focuses only those electrons with energy Ee , onto the detector. (The
energy resolution is typically 0.005 to 0.05 eV; in the particular experiment
described in Figure 4.2(a), it is 0.06 eV.) The energy E e , that is selected by
the analyzer can be varied, so that one can measure the intensity I (the
electron current at the detector) of the electrons as a function of the energy
loss E = Ee - Ee,.
The result of an actual experiment, performed with the apparatus sche-
matically depicted in Figure 4.2(a) on the diatomic molecule CO, is shown in
Figure 4.2(b).6 We see a maximum for an energy loss of Eo = Ee - E e, = 0;
i.e., a major fraction of the electron current e does not lose any energy
(elastically scattered electrons). Then there is a relative maximum of intensity
for the electrons that have lost the energy E1 (~0.26 eV), which means that
a large portion of the electrons of energy Ee lose energy inelastically to the
CO molecules. A third bump in electron intensity occurs at the energy loss
E2 = 2E 1 , and so on. As a whole, Figure 4.2 shows us that the scattered
electron current e' is a mixture of electrons that have lost one of the eight
discrete amounts of energy Eo, E 1 , E 2 , ••• , E7 at v = 0, v = 1, ... , v = 7,
respectively, to the CO molecules.
We conclude from this experimental situation 7 that the CO molecules
cannot be excited to any arbitrary energy value. Only a discrete number of
energy values are possible; i.e., the physical system has a discrete number of
energy levels. This situation is represented by an energy-level diagram
(Figure 4.3). In this particular case of the vibrating molecule, the energy-
level diagram consists of a series of equidistant levels, as can be seen from
the experimental results in Figure 4.2(b).
From Figure 4.2(b), one sees that this is only approximately true
and that the distance between energy levels actually decreases with
increasing energy. This discrepancy indicates the limitation of the
6 From G. J. Schulz, Phys. Rev. 135, A988 (1964), with permission.

7 Multiple scattering of electrons and CO molecules is negligible because the intensity of the
electron current e and the density of CO molecules are sufficiently low.
E)-----
E2 -----
llE
E 1 -----
Eo-----
Figure 4.3 Energy-level diagram of the harmonic oscillator.
harmonic oscillator as the model for the vibrating molecule. We will

discuss the corrections to this model later.
Comparing this experimental energy spectrum with the result in Equation

(3.29), we see that they agree if we take L1E = hw for the energy difference
between two neighboring energy levels. Thus the diagonal matrix elements
(cPn' H cPn) give the possible energy values of the quantum-mechanical
oscillator (e.g., vibrating molecule), apart from an overall constant. We
interpret the situation as follows: The projectile (electron) excites the
harmonic oscillator (vibrating molecule) into anyone of a discrete number
of states described by cPn' If no excitation takes place, the harmonic oscillator
remains in the ground state, described by cPo. If all the electrons that passed
through the gas had lost the same amount of energy, i.e., if there were only
one bump in the experimental curve at v = no, the ensemble of molecules
struck by these electrons would be described by one of the cPn' say cPno' In
this case the ensemble of CO molecules would be said to be in a pure state
cPno' This is often expressed by saying that "all molecules have the energy
Eno'" The results in Figure 4.2(b) show that the ensemble of molecules
considered there (i.e., those molecules that have collided with the electrons
that are scattered into the acceptance angle of the analyzer, which in the
experiment of Figure 4.2(a) is about 72° from the electron beam axis) is not
in a pure state, since the electrons have not lost one particular value Eno , but
rather a variety of energies Eo, E 1 , E 2 , .•. , E 7 . Such an ensemble of molecules
is said to be in a mixture of states, or the state of the ensemble is said to be a
mixture. This mixture can be described by the set of vectors cPo, cP l' cP2' ... ,
cPn"" and a set of numbers (relative probabilities) w o , WI"'" wn , ..• where
Wn is chosen to be proportional to the height hn of the bump corresponding
to the energy En' and they are normalized so that
L Wn = 1. (4.2)
Thus one may form a mental picture of the mixture as a collection of N

molecules, where N n = wnN is the number of molecules with energy En
in the ensemble. If it were possible to pick one molecule out of the ensemble,
Wn would be the probability that this molecule has energy En. However, a
complete dynamical description of a single specimen picked from a micro-

physical system is not possible. Wn is measured as the relative intensity of the
electron current that has lost the energy En and therefore represents a state-
ment about a large number of single specimens.
Before we introduce a concise description of the state of a mixture, we
shall give an alternate description of a pure state. To do this we shall need
some more mathematics, which we now introduce.
[An operator W is called positive if (cP, WcP) ~ 0 for all cP E Jf. A

Hermitian operator P is called a projection operator or projector iff
p 2 = P.
II.4 Experimental Data and Quantum-Mechanical Observables S9
(P - I )x
Figure 4.4 Projection operator onto a plane.
It is easily seen that the set A = {4> E Je: P4> = 4>} is a subspace
A <;; Je; for if 4>, t/J E Aand 0:, 13 E IC, then P(o:4> + f3t/J) = o:P4> + f3Pt/J
= 0:4> + f3t/J, i.e., A is closed under multiplication by a number and
under vector addition. Alternatively, A = {P4>: 4> E Je}, and hence
we write A = P Je. As an example, let Je = 1R 3 , the three-dimen-
sional Euclidean space, and let A be a plane passing through the
origin. Then for any x E 1R 3 let Px be the ordinary projection ofx onto
the plane (see Figure 4.4).
Two subspaces Al , A2 <;; Je are said to be mutually orthogonal if
for each 4> E Al and for each t/J E A2 , we have (4), t/J) = O. The set
A = {4> + t/J : 4> E Al and t/J E A2 } is called the direct sum of Al and A2 ,
and is denoted A = Al EB A2 • Similarly, given a collection Al , A2 ,
A3 , ••• , of mutually orthogonal subspaces, one can form their direct
sum, A = Li EB Ai = {Li <Pi: <Pi E A;}.
Given a projector P that projects onto the subspace A, the operator
I - P is also a projector. [Proof: Since p 2 = P, it follows that
(l - P)2 = I - IP - PI + p 2 = 1- P - (P - p 2 ) = I - P.]
I - P thus determines another subspace, which we will denote by A.l.
All vectors in A.l are orthogonal to those in A; for if 4> E Aand t/J E A\
then 4> = P4> and t/J = (l - P)t/J; using the definition (pt = P and
p 2 = P) of a projector, we have (4), t/J) = (P4>, (l - P)t/J) =
(4), P(l - P)t/J) = (4), (P - p 2)t/J) = (4),Ot/J) = O. Since for any
t/J E Je we may write t/J = Pt/J + (l - P)t/J, any projector P gives a
decomposition of Je into orthogonal subspaces, Je = A EB A.l.
For any 4> E Je, the set {0:4>: 0: E IC} is a one-dimensional subspace
of Je, called the space spanned by 4>.
Finally, given a basis {4>y} of Je, we define the trace of the operator
A by
(4.3)
It can be shown (Problem 19) that Tr A is independent of the ortho-

normal basis chosen. Tr A need not be finite. We will in general
assume that it is finite for those operators whose trace we take.

The trace has the following properties:
Tr(AB) = Tr(BA). (4.4a)
In general, Tr(A1A z ··· An) = Tr(A z A 3 ••· AnAl)'
Tr((A + B)C) = Tr(AC) + Tr(BC).] (4.4b)
We now use these new mathematical concepts to describe the harmonic
oscillator. Let
Ao be the space spanned by <Po'
Al be the space spanned by <Pl'
Az be the space spanned by <Pz,
An be the space spanned by <Pn.
Then the space Yf' is the orthogonal sum
(4.5)
Let An (n = 0, 1, 2, ... ) denote the projection operators that project on the

spaces An' Then the operator form of (4.5) is written
(4.6)
It says that every f E Yf' can be written as the sum of its components An f
in the subspaces An:
co
f= LAn!. (4.7)
n=O
This is equivalent to writing
co 00
f= L<Pn(<Pn,f) = I 1 <Pn><<Pnlf), (4.8)

n=O n=O
which is, according to (1.2. 15b), the statement that <Pn is a basis. Thus An may
also be written An = 1<Pn> <<Pn I, and (4.6) may be written
00
I = II<Pn><<Pnl. (4.9)
n=O
We will often denote the projection operator A on the one-dimensional
subspace spanned by a unit vector 1/1 by
A = 11/1><1/11. (4.10)
11.4 Experimental Data and Quantum-Mechanical Observables 61
If we apply the operator H to the arbitrary vector f and use (4.9), we have,
because of (3.29) and (3.30),
00 00
Hf = L HI¢n)<¢nlf) = L Enl¢n)<¢nlf)·
n=O n=O
As f was arbitrary, we may as well omit it and write
00 00
H = L Enl¢n)<¢nl = L EnAn· (4.11)

n=O n=O
This is called the spectral representation of the operator H. With the notion
of the trace we can express the energy values En = (¢n' H ¢n) in the form
(4.12)
PROOF. Since
ifn = y,
I\. n'f'y
A-. = {¢n
0
if n #- y,
one obtains
Tr(Hl\.n) = I (¢Y' Hl\.n¢y)
=I (¢y, H¢n)b yn = En' (4.13)
Thus the state of the harmonic oscillator in which the energy is En can be
described by the projection operator An in place of ¢n' Clearly An and ¢n
do not represent the same mathematical objects. An represents the subspace
Itn = An.Yf = {Anflf E .Yf},which is the one-dimensional subspace spanned
by ¢n (a one-dimensional subspace is often called a ray). The space Itn
contains many normalized vectors, namely, all those that can be obtained
from ¢n by multiplication with a phase factor eia , where 0 ~ II. ~ 2n. (The
set of all eia ¢ with II ¢ II = 1 is called a unit ray.)
It is generally assumed that a pure state of a physical system is described
by a projection operator A on a one-dimensional subspace (or by a unit ray)
and not by a vector. However, it is very common to call not only A but also
any vector ¢ E A.Yf a "state" or "state vector."
The result described by (4.12) is a special one because the one-dimensional
projection operator An is in the spectral representation of H. Situations do
occur where an observable A is measured in a state described by a one-
dimensional projection operator A not in the spectral representation of A
(e.g., it could be that [A, A] =f. 0). Then the "value" of the observable A in
the state A is denoted by <A) A' or just by <A), and it is determined from
measurements using the experimental set-up that defines the observable A.
<A) is obtained by the following prescription: One performs a series of N
measurements of the observable A; i.e., one either has N identical systems
on which one performs the measurement, or one performs the same measure-
ment N times on a physical system, making sure that the system is in the
same state for each measurement. Let the numbers measured for the ob-
servable A in this experiment be denoted a o , ai' ... , an' ... , and let N n be the
number of measurements that gave the result an' Then <A) is the average of
these measurements,
(4.14a)
for sufficiently large N [i.e., N must be large enough so that the right-hand
side of (4. 14a) would change only slightly if N were to be increased]. The
ratio Wn = Nn/N is the probability of obtaining the value an in one measure-
ment. Thus <A) is an experimentally determined quantity.
The operators A and A are mathematical quantities. A is the mathematical
image for the observable, and A is the mathematical image for the state. If
the mathematical image of the system (i.e., the theory of the system) is
known, then in particular one can calculate Tr(AA); i.e., Tr(AA) is a mathe-
matical quantity that follows from the theory.
The second basic assumption or postulate of quantum mechanics relates
the experimentally determined quantity <A) to the theoretically calculated
quantity Tr(AA). It is formulated now only for a pure state:
II', A pure state of a quantum-mechanical physical system is characterized

by a projection operator A on a one-dimensional subspace A:Yf (or by the
one-dimensional subspace itself). The average value <A), determined in an
experiment performed on the physical system in the state A, is related to
Tr(AA) by
<A) = Tr(AA). (4. 14b)
Tr(AA) [and because of (4.14b) also <A)] is called the expectation value of
the observable A in the state A.
For the harmonic oscillator in a pure state An' say Al (the projection
operator on the subspace spanned by ¢I)' measurements of H will always
yield the same value. Consequently
(4.15)
Hence, in the pure state AI' the probability WI for the value EI is WI = I and
all other Wi = O.
A pure state A is called an eigenstate of the observable A, or pure eigenstate
of A, if the one-dimensional space A:Yf is spanned by an eigenvector ¢ of
A. (This definition of eigenstates of A will be generalized in Section 11.6 to
include states which are not necessarily pure, i.e., not described by one-
dimensional projection operators.) Thus, the pure state Al of the harmonic
oscillator is also an eigenstate of the energy observable H, i.e., a pure eigen-
state of H. In general, as we shall see later, a physical system can be in a pure
state described by a one-dimensional space, and the measurement of some
observable A, of which this state is not an eigenstate, will still yield a set of
different numbers aI' a 2 , ... , an, .... This is the point at which quantum
mechanics deviates drastically from classical theories. We shall discuss this
matter in detail later.
We now discuss the description of a mixture rather than a pure state.
Consider the CO molecules struck by the electrons in the energy loss experi-
ments of Figures 4.1 or 4.2. If we measure the energy of the molecules N
times, we expect
No = Wo N measurements will give the value Eo,
NI = wIN measurements will give the value E I ,
N n = wnN measurements will give the value En'

because Wi is the ratio of molecules in the ensemble, with energy E i . Thus the
average value <H) of the energy operator H is
(4.16)
which is to be equal to the expectation value of H in the mixed state. We

want to write this expectation value as a trace of some operator; to do this
we introduce the positive Hermitian operator
(4.17)
where each All is the projector on the eigenspace h" of H spanned- by ¢II'
Now we calculate
Tr(HW) = L (¢Y' HW¢y)
= L L wn(¢y, HAII¢y)'
y n
Since
we have
Tr(HW) = L L w.(¢Y' H¢n)(jny
y n
(4.18)
Thus if the state of the mixture is described by the operator W defined in

Equation (4.17), then the expectation value of the energy is given by
<H) = Tr(HW). (4.19)
It is important to note that in Equation (4.19), W is not a projection operator
as would be the case for a pure state.
The statement contained in Equation (4.19) is conjectured to be the
general expression for quantum mechanics, and we formulate it as the second
postulate of quantum mechanics:
II. The state of a quantum-mechanical system is characterized by a positive

Hermitian operator W. The expectation value of an observable A, deter-
mined in an experiment as the average value <A) according to (4.l4a),
is given by
<A) = Tr(AW) if Tr W = 1
or by
<A) = Tr(A W).

Tr W
The operator W is called the statistical operator, and its matrix is called
the density matrix. The state of the system is called a pure state iff W is the
projection operator on a one-dimensional subspace.
It is customary to normalize the probabilities Wn so that
00
L Wn = 1, (4.20)
n=O
i.e., to normalize the statistical operator so that

Tr W = 1. (4.21)
Note, however, that Equation (4.20) as well as Equation (4.21) is only a
convention. Instead of wn we could have used the height of the bumps hn to
characterize the state. In that case we would get, in place of Equation (4.17),
(4.22)
n
and
It is easy to see that

W
W= Tr W·
Wand W describe the same state of the harmonic oscillator. However, if one
uses W to characterize the state, then the expectation value of the observable
H is given by
Tr(HW) . Tr(A W)
<H) = '" or m general <A) = "'. (4.23)
Tr W Tr W
Thus, whenever possible, it is more convenient to use the normalized
statistical operator for the description of the state.
The second basic assumption of quantum mechanics cannot be derived;
it can only be conjectured. It has been conjectured, in the long process of
discovering quantum mechanics, from a variety of experimental results. We
have in our above considerations made this axiom plausible for the particular
case of the energy measurement of the harmonic oscillator.
Consider the state of the harmonic oscillator given by
W = L wnAn·
n
The projection operator Ap on the eigenspace of H with eigenvalue Ep is

also an observable. Thus we can calculate the expectation value of Ap in
the state W:
(4.24)
From the properties of projection operators we know that

ApAn = 0 for p i= n,
(4.25)
AnAn = An;
consequently
(4.26)
n
(4.27)
so
(4.28)
because Tr Ap = 1, since Ap is the projector on a one-dimensional subspace.

Therefore Ap is the observable whose expectation value is the probability
of obtaining the value Ep in a measurement of the energy, i.e., the probability
of finding the system in the state Ap. In particular, for the pure state W = An
we have
<A >= Tr(A A ) = {O for p i= n, (4.29)

p p n 1 for p = n.
Generally, projection operators represent a particular kind of observable

called a property or proposition, whose expectation value tells us something
about the state.
The postulate II states how the expectation value is calculated if the
statistical operator W is known. We now discuss how the statistical operator
has to be chosen in any particular case. To do this we use our belief, which
is supported by all experimental evidence, that a repetition of a measurement
always leads to the same result, or, in other words, that all experiments
performed under the same conditions give the same experimental numbers.
It is assumed that generally a certain physical system may be in a number of
states 1, 2, ... , which in quantum mechanics are described by statistical
operators WI' Wz , ... , and that these states can be determined by the
measurements of observables A, B, .... Thus every state of the physical
system is the result of a preparation of the system. A preparation is a series
66 II Foundations of Quantum Mechanics- The Harmonic Oscillator
of manipulations with physical equipment that leads to the measurement of

an observable (or several observables). These measurements mayor may
not affect the state of the system. In classical physics the effect of a measure-
ment upon the state of the physical system is always neglected; the essential
difference in quantum physics is that here this effect is taken into account.
Thus in principle, and also in practice, not all measurements can be performed
simultaneously. Therefore one must always be able to reproduce a certain
physical state, so that it can be determined by a series of successive measure-
ments. After the measurement of an observable, the system will be in a state
that has been prepared by this measurement. This state must be described
by a statistical operator for which the calculated expectation value of the
observable agrees with the average value just measured. That is, the W
must be chosen so that the calculated value agrees with the measured value
that one will obtain in an "immediate" repetition of the measurement, i.e.,
in a measurement that is performed before the state has a chance to change
(develop in time). In fact, this principle was used when we wrote down the
state of the CO molecules. In the example of the ensemble of CO molecules,
the preparation consists of setting up the apparatus and bombarding the gas
with monochromatic electrons. This prepares the CO gas in a state that is a
mixture of energy eigenstates An' with the mixing ratios W n. The statistical
operator W has to express what has been measured on the physical system,
or what preparations have been performed. Therefore we postulate the
statistical operator to be
(4.30)
n
With this W, one calculates for the outcome of the measurement of the
energy H the numbers
Thus the original values obtained in the preparation of the state are predicted
for an immediate repetition of the measurement (i.e., a measurement before
the state W has changed), which in the present case may be any finite amount
of time later, since the situation in this experiment is stationary, but before
another measurement is made or before the setup is altered (e.g., by switching
off the electron beam).
Equation (4.30) is the third postulate of quantum mechanics for the case
that the spaces An Yf are one-dimensional. In general the space of eigenvectors
of an observable A with eigenvalue ai is not one-dimensional. Then the
eigenvalue a i is called degenerate and the spectrum {ai' a 2 , •.• } is called a
degenerate spectrum. The space Aai Yf = {fai I fai = Aai f, f E Yf} is the
space of all eigenvectors of A with eigenvalue ai . Aai is the projection operator
on the subspace Aai Yf of the space Yf. The dimension of this subspace,
dim(A a , Yf), can have any finite value or can be infinite, and can be different
for different eigenvalues a i . The space Aai Yf is called the eigenspace of the
operator A with eigenvalue a i •
After these preparations we are ready to formulate the third postulate of

quantum mechanics for the general case of an observable with degenerate
spectrum.
IlIa. Given that an observable A has been measured with the following
results
where wai gives the relative frequency with which the value a i was measured
(i.e., after N measurements the value ai occurred waiN times), then the state
of the system immediately following the measurement is represented by the
statistical operator W given by
(4.31)
where Aai are the projection operators on the eigenspaces of A with eigen-
values ai' and where dim(Aai J'l') = Tr Aai denotes the dimension of the
space Aai J'l'.
Before we discuss the general case, let us consider the particular situation
that an has been measured with the probability wn = 1 and all other
W i(i f= n) = O. Then the statistical operator is chosen as
(4.32)
[Note. The Aai are not necessarily projection operators on one-dimensional

spaces. If Aan is the projector on a one-dimensional space, the state (4.32) is
said to be a pure state.] As we shall prove below, one obtains for the expecta-
tion value
1
<A) = Tr(AW) = dim(A J'l') Tr(AAa) = an' (4.33)
an
and for the probability
1
Wa n = <Aa n ) = Tr(Aa n W) = d'1m (A J'l' ) Tr Aa n = 1. (4.34)
an
In other words, this statistical operator (4.32) predicts, for an immediate
repetition of the measurement of A, the value an with certainty. Before we
prove (4.33) and (4.34), it is convenient to present some more mathematics.
[If the space AaYf of eigenvectors of A with eigenvalue a has the

dimension dim(AaYf), then there exist dim(AaYf) linearly in-
dependent eigenvectors. One may then choose an arbitrary basis of
dim(Aa£') mutually orthonormal eigenvectors of A with eigen-

value a, i.e., a set of vectors
",1 ",2 ",K ",dim(AaX)
'+'a' 'fIa' ... , If'a' ... , 'fIa
that fulfill
(4.35)
and
(¢~, ¢~') = bKK ,· (4.36)
The basis for the whole space is given by
",1 ",2 ",dim(A(a,)X)
"Pal' 'flat' ... , o/al
(4.37)
In general, dim(Aa,£') "# dim(A aj £,). Any of the dim(A aj £,) may
or may not be infinite.
An arbitrary vector f E £' is then written, according to (1.4.4d)
dim(AaiX)
f = L ai
L1
K=
1¢~,><¢~,1f> (4.38)
This statement and the resulting form (4.42), given here without
proof, is the spectral theorem for an operator with discrete degenerate
spectrum. 8
If we apply A to the arbitrary vector f written in this form we
obtain
(4.39)
or, if we omit the arbitrary f,

(4.40)
The operator
(4.41)
K
is the projection operator on the subspace Aa, £' of eigenvectors of

A with eigenvalue ai' since
Aa,Aaj = I I
K re'
I¢~.><¢~,I¢~><¢~~I = L I¢~,><¢~jlba,aj =
K
Aa,ba,aj
8 For operators with continuous spectra one has to start with (L4.4c) instead of (I.4.4d); the
spectral theorem looks similar to (4.42) with the sum replaced by an integral, cf. K. Maurin
(1967), p. 131.
Therewith (4.40) can be written

A = L aiAa;· (4.42)
a;
This is called the spectral representation of the operator A (with

discrete spectrum). The set of Aa; is called the spectral family.
If one omits the arbitrary f in (4.38), one obtains the resolution of
the identity operator I with respect to the spectral family of the
operator A:
We now prove (4.33). We use the basis (4.37) and calculate

Tr(AAaJ = L (¢~j' AAan¢~) = L (¢~n' A¢~J
ap' K
= L an(¢~n' ¢~J = an dim(AanJf')· (4.43)

"
Thus
1
Tr(A W) = Tr(AAa ) d' (A = an'
n 1m anJf')
Equation (4.34) is established in the same way.
We now proceed with the introduction of some further mathe-
matical notions: two Hermitian operators A, B are called commuting
operators iff
AB = BA or [A, B] = O. (4.44)
Let Aa; be the spectral family of A, and P b; the spectral family of B.
Then the statement that A and B commute is equivalent 9 to
Aa;Pbj = PbjAa; for every ai' bj • (4.45a)

Thus, using (4.42), an operator A commutes with B if
APbJ. = Pb.A
J
for every b,.. (4.45b)
The spectral representation (4.42), of the Hermitian operator A

or any other operator for which (4.42) holds, can be used to define
functions of operators. 10 Let f(a;) be a function of the discrete
variable a i ; then, we define the operator f(A) by
(4.46)
a;
9 In fact (4.45a) is the precise statement of commutativity; in order to make (4.44) precise one
has to add the requirement that A 2 + B2 be essentially self-adjoint (its Hilbert space closure is
self-adjoint).
10 For the mathematical problems connected with this definition see K. Maurin (1967),
Section V.5.
Note thatf(A) might not be defined on the same subspace IfJ of

the Hilbert space £' on which A is defined; however, if f(a i ) is a
polynomial in ai' thenf(A) is defined everywhere in 1fJ.
From (4.45a) it follows that if [A, BJ = 0, then [f(A), BJ = 0.]
A pure eigenstate of A (e.g., the energy eigenstate W = An of the one-

dimensional harmonic oscillator) is a special case of (4.32).
We return now to the general expression (4.31). Comparing it with (4.42),
we recognize that (4.31) is the spectral representation of a Hermitian operator
W with respect to the spectral family {Aa) of the observable A. The
wa/dim(Aa, £') are the eigenvalues of W. For the expectation value of A, we
obtain from (4.31)
1
Tr(AW) = ~ wa , dim(Aa,Yl') Tr(AAaJ (4.47)
Inserting (4.43) into this, we obtain

<A) = Tr(AW) = I wa,a i , (4.48)
a,
as it ought to be for the expectation value, i.e., the average value in a series
of measurements of the observable A in the state of the system described by
W.
The projection operator Aa on the eigenspace of A with eigenvalue a
represents the observable "probability of a," as it should. To see this we
calculate
Here we have used

AaAa, = °for ai "# a,
AaAa = Aa'
Thus the statistical operator W of (4.31) leads to the prediction we wanted,
namely that an immediate repetition of the measurement of an observable
gives the same results as the original measurement that constituted the
preparation of the state.
If no measurement has been performed that constitutes the preparation
of a state, and no information concerning the state has been obtained, then
every pure state is assumed to be equally probable. The statistical operator
for such a state must, therefore, be taken to be proportional to the unit
operator
W=I. (4.49)
This hypothesis~that properties not affected in a measurement are

assumed to occur with the same probability~has already been incorporated
into basic assumption IlIa, i.e., into equations (4.31) and (4.32). There, no
observable that distinguishes between the different values K = 1,2, ... ,
dim(Aa,Jr) had been measured. Thus, it is assumed in (4.31) and (4.32) that
for each ai any of the pure states I¢~)<¢~,I, I¢;)<¢;,I, ... , I¢~)<¢~,I,···
appears with the same probability, namely the probability (cf. (4.34»:
w
w~=Tr(I¢~)<¢~IW)=LTr(I¢~)<¢~IAa.)-d'
l taj I J
(Aa, -./P)
1m a j Jt I I
Further, any other pure state of the space A a , Jr which is described by a vector
t/J EO A a , Jr also appears with this probability, because this probability is
w~, = Tr( It/J >< t/J IW) = ~ Tr(1 t/J >< t/J I¢~)< ¢~; I) dim(~~, Jr)
= (~ 1< t/J I¢~) 12) di~(Aa, Jr) = dim(~~, Jr)'

The last equality follows because II t/J 112 = 1 and
11t/J112 = L 1<t/J1¢~)12 = L 1<t/J1¢~)12,

aj~ K
as <t/JI¢~) = Ofora j #- ai ·
The operator A a , is the unit operator in the space A a, Jr of eigenv~ctors of
A with eigenvalue ai' which one sees immediately if one compares (4.41)
with (4.9). (4.32) then says that W is given by the unit operator in the sub-
space of eigenvectors of A with eigenvalue ai' except for the normalization,
which is chosen such that Tr W = 1. Thus the statement (4.32) is the same
as the statement (4.49) restricted to the finite-dimensional subspace Aa,Jr.
The general form (4.49) is of little practical use, because states about which
one has no information at all are of no interest. But (4.49) is frequently used
when restricted to a subspace as in (4.32).
If the system is known to be in thermodynamic equilibrium, then the
statistical operator W is chosen to be the Gibbs distribution
(4.50)
where H is the energy operator, and k and T are the Boltzmann constant
and absolute temperature, respectively. A justification of (4.50) can be found
in books on statistical physics.
One of the consequences of the basic assumption IlIa is that the possible
values of an observable that can be measured in any experiment are the
eigenvalues of that observable's operator. This means that in the process of
conjecturing the theory, the operator that represents an observable must be
chosen in such a way that all the values obtained in the experiment that measures
the observable in question are eigenvalues of the operator.
The measurement of an observable also constitutes a preparation of the
system. Generally, this means that the measurement of an observable
causes alterations in the state of the system. These statements are now
formulated in the second part of the postulate III.
IIIb. Let a system be in the state W, and let an observable B be measured

on the system. Let the eigenvalues of B be bl , b2 , b 3 , •.• , and let A b" A b2 , •.•
be the projectors on the corresponding eigenspaces. Then the state of the
system after the measurement is given by
W' = LAb, WA b,· (4.51)
b,
If this measurement is used to select a certain subensemble with a definite

value b of B, then the state is given by
1
W' = Tr(Ab WA b) Ab WA b· (4.52)
There is a special case in which the measurement of an observable does not

alter the state W; this is the case if and only if the operator B that represents
the observable commutes with W, because only then Equation (4.51)
becomes, according to (4.45b),
W' = LAb, WA b, = L Ab,Ab, W
i i
and, since
we get
W' = LAb, W = W.
i
Two observables are called compatible if the measurement of one does not
alter the state that has been prepared by the measurement of the other. The
preceding consideration then shows that compatible observables are repre-
sented by commuting operators. Observables represented by noncommuting
operators are incompatible.
If one has two pure states, WI = A", = 11/1) <1/11 and W2 = Aq, = I¢ ) <¢ I,
one can mix them to obtain a new state
0>..1.>1. (4.53)
W describes the state which is a mixture of the pure states WI and W2 with
mixing ratio A: (1 - A). It is not a pure state.
If ¢ and t/I are orthogonal to each other, i.e., Al/JAq, = 0, and one mixes
them with equal ration, A = t, then one obtains a state
(4.54)
which is described by a projection operator A. A projects on the two-dimen-
sional subspace spanned by ¢ and t/I. The probability to find the property
A"" i.e., the property of the pure state W1 , in the state W is according to (4.28)
wq, = Tr(A", W) = t Tr(A",(A", + Aq,» = t, (4.55)
as it should be.
With the two pure states A", and Aq, one can define new state vectors
(4.56)
where 0: 1 , 0: 2 are arbitrary complex numbers. All these vectors also represent
pure states. For instance one can choose 0: 1 = 1/)2 = 0: 2 , Then
(4.57)
is again a normalized vector. The projection operator
Ax = Ix><xl (4.58)
on the one-dimensional space spanned by X again represents a pure state.

The state Ax is called a superposition ofthe states A", and Aq, or the state vector
X given by (4.56) is called the superposition of the vectors t/I and ¢. To specify
the superposition of two states not only is the relative magnitude of 0: 1 and
0: 2 important, but so also is their relative phase. With the choice 0: 1
1/)2 = - 0: 2 in (4.56) one obtains the normalized vector
1
'1 = - (t/I - ¢) (4.59)
)2
which is orthogonal to X in (4.57). The projection operator
A~ = 1'1><'11 (4.60)
represents a different pure state than Ax'

In contrast to the combination (4.53) of the two states A", and Aq, which
one also calls an incoherent combination, the combination (4.56) is called a
coherent combination of the two state vectors t/I and ¢. The most general
incoherent combination of the energy eigenstates An = 1 ¢n><
¢n 1(n = 1,2, ... )
of the harmonic oscillator is (4.17) and the most general coherent combina-
tion of the energy eigenvectors ¢n is (3.27). Any change of the magnitudes
or phases of the O:n except for an overall magnitude or phase change of t/I
in (3.27) gives a vector which describes a different pure state A",. Any state
vector may be considered as the result of superposition of two or more
state vectors and this can be done in an infinite number of ways since
instead of the eigenvectors <Pn of H one can choose any other basis of the
space <D for the expansion of the vector t/I. Conversely, any two or more state
vectors can be superposed to give a new state vector. This superposition
principle is a consequence of the property (1.2.1a) of the linear space <D. If
one postulates that any vector in <D represents a pure, physically preparable
state, then any pure state can be viewed as a linear combination of an arbitrary
number of other states. Such a postulate cannot be checked because for
every vector in <D one would have to build an apparatus that prepares it.
Certainly, the energy eigenstates of the harmonic oscillator An = I<Pn>< <Pn I
are of greater importance than a pure state A", where t/I is any arbitrary linear
combination (3.27). Whether with any two energy eigenvectors <Pn and <Pm
the vector t/ln = <Pn + a<Pm' a E C also represents a physically prep arable
state is very doubtful (and certainly impossible if one considers the oscillator
as a stationary system, see Chapter XII below). However, that the linear
combination t/I 1 + t/I 2 of a certain pair of state vectors t/I 1 and t/I 2 also
represents a pure physical state is an experimentally proven fact (inter-
ference, superposition of polarized light, superposition of K-mesons).
The superpositions X and I'J span the same two-dimensional space as the
original vectors t/I and <P and the projection operator A on this two-dimen-
sional space, (4.54), can also be written as
(4.61)
Thus the mixed state W is given by
and can, therefore, as well be viewed as a mixture of the pure state, Ax and
the pure state A~ with the mixing ratios Wx = w~ = !. Whether this is of
practical importance depends upon the particular situation. If the super-
positions X and I'J represent physically preparable pure states then the state
W can be obtained by mixing the pure states A", and Aq, as well as by mixing
the pure states Ax and A~.
11.5 The Basic Assumptions Applied to the Harmonic Oscillator,

and Some Historical Remarks
We now return to the discussion of the harmonic oscillator. Let us assume

that its state has been found to be
where An is the projection operator on the space spanned by <Pn' This is the
case if a measurement of the energy always gives the result En' According
II.5 The Basic Assumptions Applied to the Harmonic Oscillator 75
to the basic assumption II, the expectation value of the momentum P in the
pure state An is given by
<P) = Tr(PW) = Tr(PA n)
= L (¢m' PAn¢m)
m
(5.1)
so
(P) = (nIPln). (5.2)
We leave it as an exercise for the reader to show that as a consequence of

(3.26) and (3.l),
(P) = o. (5.3)
Similarly the expectation value of the position Q in the pure state W = An is
(Q) = (nIQln) = O. (5.4)
We now calculate the expectation value of p 2 (i.e., the kinetic energy p 2 /2J1)
and the expectation value of Q2 (or the potential energy w 2J1Q2 /2):
(P 2) = (¢n' P2¢n)·
Using Equations (3.1), we have
p2 = _ J1wh (at _ a)2

2
so
Inserting for a and a t from Equation (3.26), we get

(P 2 ) = J1wh(n + t) (5.5)
In a similar fashion we calculate
Again from (3.1),
so
and finally
(5.6)
Although the expectation values of the observables P, Q, p2, and Q2 are

uniquely determined, it does not follow that one should expect to obtain
the same value ai for any measurement of an observable A (although such
is the case in the measurement!! of the observable H in the pure state An)'
What does follow-according to the definition of <A)-is that the average
of the values obtained in a series of many repeated measurements of A on
the state An is <A), which of course is uniquely determined. In general it
is possible that the values obtained in each measurement (a! for the first
measurement, a2 for the second, a3 for the third, etc.) are entirely different
from each other and widely dispersed.
In order to describe the dispersion of the measured values quantitatively,
let us define, for any observable A and state W, a new observable given by
(A - <A)l)2.
To see that this observable is suitable for the description of the dispersion
of the measured values a p a 2 , a 3 , ••• , we consider the expectation value of
(A - <A)l)2 in the state W, which is called the dispersion of A in the state W:
disp(W)A == «A - <A)wl)2)w
= Tr(A2W) - 2 Tr(A<A)w W) + Tr«A)fv W)
= Tr(A2W) - (Tr(A W))2 = <A2)W - <A)fv (5.7)
(unless otherwise stated we always assume that Tr W = 1); thus
(5.8)
For the special case of the observable H for the harmonic oscillator in the
state An we obtain
disp(A n) H = Tr(H2 An) - (Tr(H An))2
= (¢n' H2¢n) - E;.
However, since ¢n is the eigenvector of H with eigenvalue En' we have

H2¢n = HEn¢n = EnH¢n
= E;¢n;
thus
disp(A n ) H = E; - E; = O.
This shows that the dispersion of H in the state An is zero. More generally,
the dispersion of an arbitrary observable A in a pure eigenstate A of A is
zero. However, the converse ofthis statement is not true; i.e., if the dispersion
of an observable A is zero, then the state W need not be a pure eigenstate of
A (in particular, it need not be a pure state). For this reason, the notion of
eigenstate will be generalized.
11 We remark that by measurement we always mean an idealized measurement in which the

measuring apparatus does not lead to uncertainties; i.e., the experimental error is zero.
II.5 The Basic Assumptions Applied to the Harmonic Oscillator 77
Let A be an observable with degenerate spectrum. A state W is called an

eigenstate of the observable A if W can be represented as
W = L wl<l<p~)(<p~I, (5.9)
I<
where wI< ;;::: 0, LI< W" = 1 (W is normalized), and <p~, K = 1,2, ... , dim AaYf
is any basis of the eigenspace Aa Yf of A with eigenvalue a.
In Section 11.6 it will be shown that the eigenstates of an observable are
precisely those states in which the dispersion of A is zero. A special case of
an eigenstate is (5.9) with w" = w = (dim AaYf)-l for which W is given by
1
W = Aa Tr(Aa)' (5.10)
The energy eigenstate An for the harmonic oscillator is not an eigenstate

of either of the observables P and Q, so its dispersion is not zero. The disper-
sion is easily calculated:
disp(A n ) P = (P 2) _ (P)2.
From Equations (5.3) and (5.5),
(P 2 ) _ (P)2 = lil1w(n + t). (5.11)
Similarly
· Ii
d ISP(A)
n
2 2
Q = (Q ) - (Q) = - (n
I1 W
+ 2)
1
(5.12)
The square root of the dispersion is called the uncertainty of A in the state
W, and is denoted by ~A:
~A = Jdisp A. (5.13)
From Equations (5.11) and (5.12) it follows that
~P ~Q = Ii(n + t). (5.14)
Thus for any pure energy eigenstate of the harmonic oscillator,
~P ~Q ;;::: lit. (5.15)
This relation is true in general, as we shall see later, and is a consequence of
the Heisenberg commutation relations [P, Q] = (Ii/i)!. It is called the
Heisenberg uncertainty relation.
As shown above for the particular case of the pure eigenstates of the
observable H, the dispersion of an observable in an eigenstate of that ob-
servable is zero. Physically, this means that a measurement of the observable
on an eigenstate gives one value (which must be an eigenvalue of the corre-
sponding operator) with certainty, i.e., with unit probability.
Now consider the mixture W of the energy-loss experiment described by
(4.17). By definition
(5.16)
Using (4.1S) and the corresponding expression for <H 2 ), we have
<H2) - <H)2 = ~ E;wy - (~WyEyr

(5.17)
!1E is called standard deviation or root-mean-square (Lm.s) deviation. For

the last inequality we have used the Cauchy inequality for sequences:
(5.1S)
which is a consequence of(I.2.7). Thus disp(W) H in a mixture is always greater

than or equal to zero. It is equal to zero if
and it deviates from zero according to the number of and values of Wy that
differ from zero. Thus disp H, and therewith /)']i, is indeed a quantitative
expression for the uncertainty in the measurement of the observable H. From
the first line in Equation (5.17) we see that the uncertainty !1H of the ob-
servable H corresponds in the measurement of this observable to the standard
deviation or Lm.s. deviation of the series of measurements described above
Equation (4.16).
It is no surprise that disp H or !1H was greater than zero for the mixed
state W of CO molecules in the collision chamber, since that was a mixture
of CO molecules of different energies. The reason that the outcome of the
measurement of H in W is not fully determined is that not enough prepara-
tions had been performed to separate different systems, or more precisely,
to separate systems in different states, from each otheL
The harmonic oscillators in the collision chamber can be separated, at
least in a gedanken experiment, into eight sets, each one being a set of N n
molecules having the same energy En' i.e., such that a measurement of H
on the nth such set gives with certainty the value En' The state of each set is
then An' It is impossible to further separate (even in a gedanken experiment)
any of these sets into two or more subsets of different systems or systems in
different states. This is the reason for calling An a pure state. If we now per-
form the measurement 12 of an observable on the system in a pure state, one
would expect from classical considerations that this should lead with
certainty to a well-defined number, as a set in a pure state consists of only
one species. If we choose for this observable HorN, this is indeed the case.
However, if we choose P or Q, we see from (5.11), (5.12) that it is not. Even
in pure states the outcome of the measurement of an observable is in general
not uniquely determined. Thus statistics cannot in principle be eliminated
from quantum mechanics; systems in the "same pure state" do not give
12 An idealized measurement; see footnote 11.

11.5 The Basic Assumptions Applied to the Harmonic Oscillator 79
identical values in a measurement. All that can be said is with what proba-
bility Wn a certain value an can be expected in the measurement of an ob-
servable.
Let us illustrate this with a gedanken experiment. We consider the measure-
ment of the observable Q in the state An' and ask for the probability with
which we can expect the value x in a measurement of Q. By definition, the
expectation value of an observable A is
(5.19)
where a i are the eigenvalues of A, and Wi are the probabilities for finding the
values ai in a measurement of A. We calculate <Q) as follows:
<Q) = Tr(QAn) = I <iIQAnl i), (5.20)
where I/Ii = Ii) is any complete basis system and we have used the following
notation:
(rP, 1/1) = <rPll/I) or (I/Ii' 1/1) = <ilj),

(5.21)
(rP, AI/I) = <rPIAII/I)·
For Ii) we may take any complete basis system; let us therefore choose the
system of eigenvectors of the operator Q, which we denote by Ix), where x
is the eigenvalue of Q:
Qlx) = xix). (5.22)
Then
<Q) = I <xIQAnlx)
x
= I x<xIAnlx)
x
= I x<xln)<nlx).
x
The last step follows from writing An = In)<nl.

We introduce new notation, defining
rPnCx) == <xln). (5.23)

Then
(5.24)
x
If we now compare this with the general expression (5.19) for the expectation
value of an operator, we conclude that
(5.25)
80 II Foundations of Ouantum Mechanics-The Harmonic Oscillator
is the probability for obtaining the value x in the measurement of the ob-
servable Q for an oscillator in the state An' ¢n(x) is called the wave function
of the state with energy En' Thus we have found that the probability for
finding the value x for the position operator Q is the absolute value squared
of the wave function. It was in this form that the probability interpretation
was originally introduced in quantum mechanics by M. Born [influenced by
an earlier suggestion of Albert Einstein (1916)].
As we shall discuss later, the Lx
is in fact an integral, and wn(x) = 1 <x 1 n) 12
is a quantity of dimension cm - 1, i.e., a probability density. The probability
is the integral of wn(x) over a certain interval.
We have derived the fact that 1 ¢(xW describes the probability of finding
the physical system at x from the basic assumption II. Historically the events
took place in reverse order, and II is a generalization of the discovery by
Max Born in June 1926 that the wave function ¢(x), which-as we shall see
in Section II. 7~is a solution of the Schrodinger equation, is connected with
probability.
Born was at that time investigating the collision process between free
particles and an atom, using the formalism of Schrodinger's wave mechanics,
which for this kind of process he found superior to the matrix mechanics of
Heisenberg, Born, and Jordan. However, he could not accept Schrodinger's
interpretation that ¢(x) represents matter waves. Recalling these events,
Born said in his Nobel Prize Lecture in 1954: "On this point [Schrodinger's
matter-wave interpretation] I could not follow him. This was connected
with the fact that my Institute and that of James Franck were housed in the
same building in Gottingen University. Every experiment by Franck and
his assistants on electron collision appeared to me as a new proof of the
corpuscular nature of the electron." He calculated the formula for the
scattered wave in the Schrodinger formalism (Born approximation 13) and
concluded that if this formula is to allow for a corpuscular interpretation,
then the most natural interpretation of ¢(x) is that 1¢(x) 12 describes the
probability density for detecting the corpuscle. He summarized his con-
clusion in 1926: "The motion of particles conforms to the law of probability
but the probability itself is propagated in accordance with causality."
Born's interpretation was soon generalized to include other observables
in addition to the position. And the time was ripe for the general formulation
of quantum mechanics. The starting point for this was the discovery that
the canonical transformations can be carried over to matrix mechanics and
wave mechanics (F. London, May 1926; P. Jordan, July 1926). In this way
the canonical transformations were recognized as transformations of an
abstract space. Transformations of objects with continuous indices were
introduced by Dirac and Jordan in December 1926. The mathematical
elaboration of the London-Jordan-Dirac transformation theory was under-
taken in Gottingen by David Hilbert together with L. Nordheim and John
von Neumann (1926-1927). Notions like state and observable evolved.
13 See Section XIV.S below.

JI.6 Some General Consequences of the Basic Assumptions 81
Subsequently, von Neumann developed an axiomatic theory of Hilbert

spaces and established the association between physical states and Hilbert-
space vectors as well as the correspondence between observables and linear
operators. He also introduced the statistical operator.
Parallel to the mathematical formulation went the development of the
physical interpretation, which was discussed principally in Copenhagen
(1926-1927). Whereas Bohr wanted to make the wave-particle duality the
starting point of the physical interpretation, Heisenberg relied upon the
transformation theory. His basic approach was to regard something as
defined if it is measurable. In his famous gedanken experiment in which
he discussed the position measurement of the electron with a microscope,
he explained the uncertainty of position and momentum (5.15), and there-
with the probabilistic nature of quantum-mechanical predictions, as the
effect of the measurement apparatus. Object and observer have no indepen-
dent reality; the measurement affects the state of the object (basic assumption
III).
While Heisenberg derived his uncertainty principle from the transforma-
tion theory, Bohr developed his conception of complementarity from the
wave-particle duality (see Section 11.9): There exist complementary proper-
ties-like position and momentum-and the exact measurement of one
precludes the possibility of obtaining information of the other. Properties
are not actualities; they are only possibilities for the physical system. These
developments formed the basis of the so-called Copenhagen interpretation
of quantum mechanics. 14 Its physical content is formulated in the basic
assumptions of this chapter.
Von Neumann's Hilbert-space formulation could not accommodate
continuous eigenvalues. Since that time, stimulated by the Dirac formalism,
a new branch of mathematics-the theory of distributions-was developed
and for the first time systematically presented by L. Schwartz (1950-1951).
On this basis I. M. Gelfand and collaborators introduced around 1960 the
rigged Hilbert space. Its use for quantum mechanics was suggested around
1965.
11.6 Some General Consequences of the Basic Assumptions

of Quantum Mechanics
After this detailed discussion of the case of the harmonic oscillator, let us
investigate some general consequences of the basic assumptions of quantum
mechanics. Let us consider an arbitrary state W of an arbitrary micro-
physical system and see what we can predict for the measurement of the
14 The interpretation of quantum mechanics suggested by Born and established by Heisen-

berg and the Copenhagen school represented such a drastic departure from the well-established
ideas of their time that decades had to pass before it could gain general acceptance, and even
today new attempts are being undertaken to return to a deterministic interpretation. Born's
statistical interpretation of quantum mechanics was one of the most far-reaching contributions
of physics to contemporary thinking.
values of some arbitrary Hermitian observables A, B, ... ofthis system. These

predictions are contained in the following two statements, which we will
prove below.
1. The dispersion of any observable A in an arbitrary state W fulfills
disp(w) A ~ 0 (6.1)
with
disp(W) A = 0
if and only if W is an eigenstate of A.
2. The uncertainties of two observables A, B in an arbitrary state W
fulfill the uncertainty relation
dA dB ~ !I<[A, BDwl. (6.2)
A special case of this uncertainty relation is the Heisenberg uncertainty
relation; from
[P, Q] = ~I I and (--:-Ii)

I
I Ii Tr(WI) Ii
= --:-
I
= --:-
I
it follows that
(6.3)
for any state W. The equality sign in the Heisenberg uncertainty relation
holds if W = Ao , the ground state of the one-dimensional harmonic
oscillator, and for certain other pure states as well (cf. Problem 4e).
The uncertainty relation is a very general property of two operators, and
just depends upon the assumption that the observables are represented by
linear operators in a linear space. It says that unless A and B are compatible
observables, there are always states for which at least one of the observables
cannot be measured with exact accuracy (since for every operator C = irA, B]
there is always a W such that
I<C)wl = ITr(WC) I i= 0).
In the case that C (or - C) is positive definite (as, e.g., for A = P and B = Q),
ITr(WC) I > 0 for every state W, and therefore the two observables A and
B can never be measured simultaneously with unrestricted accuracy. If the
system has been prepared in an eigenstate of one of the observables A,
dA = 0 (i.e., A has been measured to give accurately one value), then
according to (6.2) the values for B are completely uncertain. This is a con-
sequence of the previously discussed fact that in quantum physics a measure-
ment changes the state, and the measurement of one observable interferes
with the measurement of another observable unless these two observables
are compatible, i.e., are represented by operators that commute with each
other. Thus only compatible observables can be simultaneously measured,
and only for compatible observables do there exist common eigenstates.
n.6 Some General Consequences of the Basic Assumptions 83
PROOF OF STATEMENT 1. A variant of the spectral theorem (4.42) states that

every statistical operator can be written in the form
(6.4)
with real numbers Wi' and one-dimensional projection operators Ai which project
onto spaces Ai£ spanned by single vectors 4>i such that the vectors 4>1' 4>2"'"
together form an orthonormal basis:
(6.5)
The numbers Wi satisfy·
(6.6)
[which follows from the positivity of W since by (6.4) and (6.5) we have Wi =
(4)i' W4>;)], and
I Wi = Tr W= 1. (6.7)
Since Tr(BW) for an arbitrary operator B can be evaluated with any basis, we
choose the particular basis {4>J from (6.5) and obtain
Tr(BW) =I (4)j, B I WiAi4>j) =I Wi(4)i' B4>;), (6.8)

) , ,
where we have used Ai4>j = bij 4>i' Taking B = I in (6.8) leads to (6.7) because
(4)i' 14>i) = (4)i' 4>i) = 1 for all i.
From the definition (5.7), we have
disp(w) A = <CA = Tr«A - ex1)2W)
- ex1)2)W (6.9)
where ex = <A)w = Tr(AW). Thus, taking B = (A - exl)2 in (6.8) yields
disp(w) A =I Wi(4)i' (A - ex1)24»
(6.10)
Therefore disp(W) A :2 0, since Wi :2 0 by (6.9) and obviously II(A - ex1)4>iI1 2 :2 0 for

all i. Furthermore, disp(w) A = 0 implies
(6.11)
For a given i, this in turn implies that either Wi = 0 or II(A - ex1)4>iI1 2 = O. If Wi = 0,

then WiAi is also zero, and this term can be omitted from the sum in (6.4). Thus the
sum in (6.4) can be taken to run only over those values of i having Wi > O. For the
values oU which remain, II(A - exl)4>iI1 2 = 0 is required by (6.11); i.e.,
(6.12)
Therefore, if disp(w) A = 0, then all 4>i for the remaining values of i must be eigen-
vectors of A belonging to the same eigenvalue ex. But then the state Win (6.4) describes
an eigenstate of A as defined in (5.9).
The converse, i.e., that disp(w) A = 0 for W given by (5.9), is proven by a straight-
forward calculation (see Problem 10). 0
PROOF OF STATEMENT 2. For two operators A and B we define

h(A, B) == Tr(WAtB) = Tr(WI(AtB» = h(J, AtB) = <AtB>. (6.13)
One easily shows from (4.3) and (4.4) that

(a) h(A, A) ~ 0_,_
(b) h(A, B) = h(B, A),
(c) h(aA, B) = llh(A, B),
(d) h(A + C, B) = h(A, B) + h(C, B),
i.e., h(A, B) defined by (6.13) fulfills all the conditions of (1.2.5) that define a positive
Hermitian form. Since h(A, B) is a positive Hermitian form, the Schwarz inequality
(1.2.7) holds:
1h(A, B) 12 ::;; h(A, A)h(B, B). (6.14)
From this, statement 2 is easily proven (we assume A, B to be Hermitian):
1<[A, BJ> 12 = Ih(J, [A, BJW = Ih(A, B) - h(B, A)1 2 ::;; 4h(A, A)h(B, B)
so if we replace
A -> A - rxI, rx = Tr(WA),
B -> B - f3I, f3 = Tr(WB),
we obtain
h(A - rxI, A - rxI)h(B - f3I, B - f3I) ~ tlh(J, [A - rxI, B - f3IJ)1 2
or (for A and B Hermitian)
disp A . disp B ~ tlh(1, [A, BJ) 12 = tl <[A, BD 12. (6.15)
The square root of this gives (6.2), and therefore statement 2 has been proven. 0
11.7 Eigenvectors of Position and Momentum Operators;

the Wave Functions of the Harmonic Oscillator
In this section we shall investigate the properties of the wave functions.

We will begin by examining the properties of the operators P and Q; in
particular, we wish to know what are the possible eigenvalues x of Q. That
is, for which values x can
Qlx) = xix) (7.1)
be fulfilled? Then we wish to find the transition coefficient (x In) which

transforms between a vector In) = <Pn (an eigenstate of Hand N) and the
vectors Ix) (eigenstates of Q):
In) = L Ix)(xln). (7.2)

x
II.? Wave Functions of the Harmonic Oscillator 85
From (3.1) and (3.26) it follows that
Qln) = J2:W (a + at)ln)

= J h (Jnln - 1)
2p.w
+ In+lln + 1)). (7.3)
Taking the scalar product of this equation with Ix), we obtain
(nIQlx) = J2:W (In(n - 11x) + In+l(n + 1Ix)). (7.4)
On the other hand, taking the scalar product of (7.1) with In), we obtain
(nIQlx) = x(nlx). (7.5)
Comparing (7.4) with (7.5) yields
x(nlx) = J2:W (In(n - 11x) + In+l(n + 1Ix)), (7.6)
or, denoting n + 1 = m,
jm(mlx) = J2~W x(m - 11x) - ~(m - 2Ix). (7.7)
Since (7.3) is valid for n = 0, 1,2, ... , (7.7) is valid for m = 1,2, 3, .... For
n = 0 (i.e., m = 1), we obtain instead of (7.3)
QIO) = J2:W JO+lIO + 1) = J2:W 11); (7.3a)
and instead of (7.7),
v'11
1 (llx) = J2P.W
-h- x(Olx). (7.7a)
Thus we see that (7.7) is a recurrence relation for (mix); if (Olx) is known,
we can determine (llx) by (7.7a) and then determine (2Ix) by (7.7). With
(11 x) and (21 x) we can determine (31 x) by (7.7), and so on.
To find out what the transition coefficients (m Ix) are, we introduce
(7.8)
and
= ~(nlx)
fn(Y) - v' 2 n! (0 Ix)' (7.9)
which is defined for all x with (Olx) -=f. 0 (if (Olx) = 0, then by (7.7) and
(7.7a) (nix) = 0 for all n). Then from (7.7) it follows that
J2m:! !m(Y) = J2m l(~ _ 1)! Y!m-I(Y) - 2m ~~ ~ 2)! !m-iy),

(7.10)
or
!m(Y) = 2Y!m-I(Y) - 2(m - 1)!m-2(Y)' (7.11 )

From (7.7a) we have
(7.12)
and from (7.9)
(7.13)
Equations (7.11), (7.12), and (7.13) are the recurrence relations for the
Hermite functions and have solutions for any complex number y.15 Thus
for any complex value x there is a solution (n Ix) of the recurrence relation
(7.7). For physical reasons we need to consider only real values of x, because
x is the value that is obtained in the position measurement and must be real.
For real values of y, the solutions !m(Y) of (7.11), (7.12), (7.13) are the Hermite
polynomials:
(7.14)
Thus from (7.9) we may obtain the transition coefficient (n Ix) for every real
value of x for which (0 Ix) is defined. We restrict ourselves to those solutions
of (7.7) for which (0 I x) is finite, because I (0 I xW, the probability for obtain-
ing the value x in a measurement of Q in the ground state Ao , is assumed to
be finite.
Combining (7.8), (7.9), and (7.14), we have
(nix) = ~ (0 IX)Hn(if: x) (7.15)
for - 00 < x < + 00.
[The set (1/IIx) = Ln Ctn(<Pnlx), considered for a fixed value x as a

function of 1/1, is an antilinear functional over the space of the 1/1
in Equation (3.27) [i.e., it fulfills Equation (1.7.1)]. Thus Ix) is a
continuous functional, cf. Section 1.7]. As was mentioned in Chapter
I, the set of continuous linear functionals on a space <1>, which we
denote by x , is in general larger than the space <1>, except for the
15 Cf. Appendix of Section 1.6 or N. N. Lebedev, Special Functions and Their Applications,
Prentice Hall, Englewood Cliffs, N.J., 1965, Section 10.4.
II.7 Wave Functions of the Harmonic Oscillator 87
Hilbert space Yf, where Yfx = Yf (Frechet-Riesz theorem.)

Taking into account (3.28), we obtain <l> c Yf c <l> x, the Gelfand
triplet or rigged Hilbert space for the harmonic oscillator. The Ix)
that obey (7.1) are elements of <l>x and are the generalized eigen-
vectors of the operator Q; the set
Ix) -00 < x < +00
is a generalized basis system. The spectrum of Q is continuous and

is the whole real axis IR. The spectrum of the operator N, (3.2), or
H, (3.7), is discrete. As x can be any real value, the formal sum in
(7.2) is actually an integral. (7.2) is a special case of the nuclear
spectral theorem, according to which every cP E <l> can be written as
cP = fdXIX)<XICP) = fdXIX)CP(X), (7.16)
as was explained in detail in Section 104. Here we denote the gener-

alized eigenvectors of Q,
Qlx) = xix), (7.17)
by Ix) instead of Ix) to emphasize the difference between them and

the eigenvectors belonging to the discrete spectrum, I n). If we take
the "scalar product" of cP in (7.16) with Ix'), or more precisely the
functional Ix') at the element cP, we have
(7.18)
<x' Ix) is therefore that mathematical object which by integration

(7.18) maps the "well-behaved" function cp(x) = <xlcp) into its
value at the point x': cp(x') = <x'i cp). It is the Dirac b1unction or
b-distribution
<x'lx) = b(x' - x), (7.19)

so (7.18) is written
cp(x') = <x'lcp) = f
dx b(x' - x)<xlcp)
= f
dx b(x' - x)cp(x). (7.20)
In the transition from the symbolic (7.2) to (7.16),
L Ix)(xl
x
was replaced by f dx Ix)<xl·
Consequently, the vectors Ix) and the generalized vectors Ix) must
have different dimensions in order that
Ix)(xl and dx Ix)<xl
may have the same dimension.
The Ix) <x I dx are the analogues of the projection operators;
however, they do not project on any vector of the Hilbert space Yf
(or <1», because the Ix) are not elements of these spaces. Projection
operators are J,h dx I x) <x I where the integration extends over a
finite interval ,1x. As in the case for operators with discrete spectra,
one can define functions of the operator Q with continuous spectrum
by
f(Q) = fdx f(x)lx)<xl,
where f(x) is a function defined on the spectrum of Q with

Jdx f(x)ltfJ(x)1
2 < 00 for all tfJ E <1>.]
We now return to the evaluation of the harmonic-oscillator function and

calculate (Olx). We write ¢n = In) in the form (7.16),
In) = f dx Ix)<xln), (7.16')

and take the scalar product of this with ¢m'
c5 mn = (¢m' ¢n) = (min) = f dx (mlx)<xln). (7.21)
We shall make use of

<xln) = (nix). (7.22)
[If In) = ¢n and Ix) = Xboth were vectors in the Hilbert space then
(7.22) would be the condition (1.2.2b) of the definition of the scalar
product. However (n Ix) is the functional Ix) E <1> x at the element
In) E <1> and <x In) is a functional In) E <1> x x = <1> at the element
Ix) E <1> x , so (7.22) is not obvious. But it can be shown that (7.22) is
generally true for transition coefficients between a discrete basis and
a continuous basis. Hereafter we shall drop the typographical
distinction I ) as opposed to I ), and shall use generally I ), unless
we want to stress the distinction between generalized and ordinary
vectors.]
We insert (7.15) into (7.21) and obtain
(7.23)
II.7 Wave Functions of the Harmonic Oscillator 89
If we compare this with the orthogonality relations for the Hermite poly-
nomials,16
f -
dye
y2 _ {O
Hm(y)H.(y) -
for m oF n,
'2.;: f'
n. V n lor m = n,
(7.24)
we find
(7.25)
Thus, up to an arbitrary phase factor (which we choose to be unity),
<Olx) = (~~r/4e-<JlWI2h)X2. (7.26)
With this and (7.15) we obtain the transition coefficients <nix) between the
x- and n-bases, i.e., the harmonic-oscillator wave functions 4>.(x):
<nix) = (IlW) 1/4_1_ H ( ~ x)e-<JlWI2h)X2 = <xln) = 4> (x). (7.27)

nh ~. -Jh ·
A few of the wave functions are plotted in Figure 7.1.
We now repeat the procedure for the operator P that we have already
gone through for the operator Q. The eigenvectors of P will be denoted
by Ip):
Pip) = pip) (7.28)
The action of P on In) (cf. Problem 13) is
Pin) = -i JhllW
-2-(a - at)ln)
= -iJh~W (Jnln - 1) - In+lln + 1»). (7.29)
16 Lebedev, ibid., Section 4.13.
",.(xl
"- 0
Figure 7.1 Wave functions of the harmonic oscillator [from Ludwig (1954), with
permission ].
If we take the" scalar product" with Ip) and use (7.28), we obtain
p(pln) = -iJh~W (In(pln - 1) - jn+l (pin + 1», (7.30)
or
p(nlp) = iJh~W (In(n - lip) - jn+l (n + lip». (7.31)
If we introduce the new quantities (nip) defined by
(nip) = in(nlp), (7.32)

then
i(n - lip) = ii n- 1 (n - lip),
(7.33)
-i(n + lip) = _iin+l(n + lip),
so (7.31) may be written
p(nlp) = Jh~W (In(n - lip) + jn+l(n + lip». (7.34)
We see that this is exactly the same recurrence relation as in (7.6), with
xj/lw/h replaced by p/~. Thus by the same argument as for (nix), we
find [using (7.32)J that
(nip) = in - - 1)
(n/lwh 1/4 1 (1 )
- - - Hn - _ p e- p2 / 2 I'wh. (7.35)
j2 n n! jh/lW
We have seen that the eigenvectors In) of the energy operator H for the
harmonic oscillator have the very particular property that the transition
coefficients (7.27) between these vectors and the x-basis are the same as the
transition coefficients (7.35) between these vectors and the p-basis except
for a phase factor.
By the same argument as above for the operator Q, we conclude that the
spectrum of P is continuous,
spectrum P = {pi - 00 < p < oo}, (7.36)
and the Ip) are generalized eigenvectors.

The transition coefficients (p In) in
In) = fdPIP)(Pln) (7.37)
are called the wave functions in the momentum representation and are denoted
¢n(P) == (pin). (7.38)

11.7 Wave Functions of the Harmonic Oscillator 91
The probability density of finding the value P in a measurement of the

momentum P in the energy eigenstate An of the oscillator is wn(p) =
I(p In) 12. Also, for any arbitrary vector cp the transition coefficient
cp(p) = (plcp) (7.39)
in
cp = f
dp Ip) (plcp) (7.40)
is called the momentum-space wave function or the wave function in the

p-representation for the state cp.
So far, we have obtained the matrix element of Q in the x-representation,
(nIQlx) = x(nlx); (7.41)
or more generally, for every well-behaved vector cp,
(cpIQlx) = x(cplx)' (xIQlcp) = x(xlcp)· (7.42)
Likewise, we have obtained the matrix element of P in the p-representation,
(nIPlp) = p(nlp); (7.43)

or more generally,
(cpIPlp) = p(cplp), (pIPlcp) = p(plcp)· (7.44)
We now want to calculate the matrix elements of P in the basis of

generalized eigenvectors of Q and the matrix elements of Q in the basis of
generalized eigenvectors of P. We do this in two steps: First we find out
what the symbol (x Ip), the transition matrix element between the x-basis
and the p-basis, represents, and then we determine (x IPin).
We begin by taking the scalar product of (7.16') with Ip) [or, more pre-
cisely, we will consider cPn as a functional at the generalized eigenvector
Ip) E x , P E spectrum P, and use (7.16')]:
(pin) = f dx (plx)(xln); (7.45)
and then we take the scalar product of (7.37) with (xl [or more precisely,
consider cPn as a functional at the generalized eigenvector
X E spectrum Q,
and use (7.37)J:
(xln) = f dp (xlp)(pln). (7.46)

In (7.45) and (7.46), <xln) and <pin) are given by (7.27) and (7.35), respec-
tively. The Hermite polynomials have the property [Equation (I.A.17)r 7:
ine-~2/2Hn('1) = f -
+OO
00 v' 21t
i~~
d¢ e~ e-~2/2Hn(O' (7.47)
Inserting (7.35) and (7.27) into this relation, it follows that
<n Ip) = f
-
+ 00
00
eixp / h
dx M:::i. <n Ix),
v' 21th
(7.48)
or, taking the complex conjugate,
<pin) = J dx
e- ixp/h
M:::i. <xln).
v' 21th
(7.49)
Comparing (7.49) with (7.45), we find that the <p I x) are given by
<pix) = ~1~ e- ixp / h• (7.50)

~
In the same way one obtains from (7.46) and (7.47)
<xlp) = ~1~ eixp / h• (7.51)

~
(7.50) and (7.51) together give
<xlp) = <pix). (7.52)
It is now simple to calculate the matrix element of P in the basis of gener-

alized eigenvectors of Q, using (7.51):
<xIPlep) = J dp p<xlp) <pi ep) = J

dp p Z <plep)
= f dPihddx <xlp)<plep) ihdf

dx dp <xlp)<plep)· =
Thus
hd
<xIPlep) = -;--d <xlep)· (7.53)
I x
In the same way, using (7.50), one obtains
h d
<pIQlep) = - -;--d <plep)· (7.54)
I P
17 Lebedev, ibid., Section 4.12.

n.7 Wave Functions of the Harmonic Oscillator 93
Each vector in the linear space can be fully characterized by its components
with respect to a basis system. In Sections 1.4 and 1.5 it was discussed in
detail that in the same way as one represents the vector x in the three-
dimensional space by its component Xi with respect to the three basis vectors
e i , according to x = L~= 1 eix i , one can represent the vector cp by its com-
ponents <n Icp) with respect to the discrete basis CPn = In) of eigenvectors of
H according to cP = 2::=0 In)<nlcp) [Equation (4.8)]. This is also true if
the basis is a continuous basis, according to
cp = fdX,x)<X'CP). (7.16)
Thus to the vector cP corresponds the function <x Icp) = cp(x) and to the
vector Pcp corresponds the function <x IP Icp). Equation (7.53) then states
that in the realization ofthe space of vectors cp by the space of wave functions
< x Icp) = cp(x), the momentum operator is realized by the differential
operator times fiji, and (7.42) states that the position operator is realized
by the operator of multiplication with the number x. This realization is
called the Schrodinger representation.
[It should be noted that the Schr6dinger representation could not

be derived from the commutation relation (2.1a) alone, but required
an additional assumption: that the operator H in (2.1 b) has at
least one proper eigenvector. This assumption is one of several
equivalent forms of the additional requirement. Other forms are that
p 2 + Q2 be essentially self adjoint or that the representation of the
Heisenberg algebra be integrable to the representation of the Weyl
group (Weyl's form of the canonical commutation relation).]
The Schr6dinger representation of the energy eigenvalue equation
(7.55)
is the time-independent Schr6dinger equation which we shall derive now

for the harmonic oscillator. The energy operator for the oscillator was
given by (2.1b):
(7.56)
Let us take the matrix element of H between <x I and In):
(7.57)
We compute
<xlp 2 In) = L <xIPln')<n'IPln)
n'
nd
= L ~-d
n'I X
<xln')<n'IPln)
= ~ dd L <xln')<n'IPln)
I X n'
nd
= ~- <xIPln)
I dx
= (in)2 dx2
d2
<xln). (7.58)
Moreover, from (7.1),

(7.59)
Inserting these in (7.57), we have for the matrix element of the energy oper-
ator
(7.60)
Therefore, in the Schrodinger representation the eigenvalue equation (7.55)

IS
(7.55')
Summarizing, we have seen that in the x-representation, the operator P

is given by - in d/dx,
d
P <---> -in- (7.61)
dx'
and the energy operator by
11.8 Postulates II and III for Observables with

Continuous Spectra
The description of the basic assumptions II and III in Section 11.4 was
given in terms of operators with discrete spectra. In this section we shall
give the description for the case that the observables have continuous
II.S Postulates II and III for Observables with Continuous Spectra 95
spectra. With the help of the formalism introduced here we shall be able to
discuss in more detail the problems connected with the measurement of
position and momentum, the subject of the next section.
The statistical operator of a pure state,
W = Abo = Ibo><bol,
where Abo is the projection operator onto the one-dimensional subspace
spanned by the (proper) eigenvector Ibo>of an observable B, is written in
an arbitrary basis system {<Pn} as
(8.la)
mn
where
(8.lb)
If {<Pn} is the basis system {<Pb = Ib>} in which B and therefore W is diagonal,
then
C b = <<Pblbo> = <blb o> = (jbbo' (8.lb')
For a mixture the statistical operator is
(8.2)
mn
where Wmn are the matrix elements of W in the basis {<Pn}; the collection
{w mn } is the density matrix.
In the case of a continuous spectrum Equation (8.la) has to be replaced
by
W = f
dx' dx" f(x')j(x")lx'><x" I, (8.3a)
where {I x>} is the basis system of generalized eigenvectors of an observable,

say Q, with a continuous spectrum. The (well-behaved test) function f(x)
of the continuous variable x for the pure state W = Abo is given by
(8.3b)
Here we have used the second notation to emphasize that Ix> is a generalized
eigenvector and Ibo) is a proper eigenvector. For a mixture the statistical
operator is in analogy to (8.2):
W = f dx' dx" F(x ', X")IX'><x"l, (8.4)
where F(x', x") is a function of the two continuous variables x' and x", the
generalized eigenvalues of Q.
Let us now assume that the observable B that was used to prepare the
state W is Q, and that we want to prepare and describe a pure state. Naively
one would want to choose
in analogy to the discrete case (8.1 b'), so that

W= Axo = IXo><xol.
However, it does not make sense to ask for a measurement of a particular
point of an observable with a continuous spectrum. Any measurement
apparatus, a macroscopic system, is of finite size, not infinitesimally small.
Counters have finite extension and need a finite amount of energy; slits-as
narrow as they may be-have finite extension. Therefore one cannot ask
for a state of a physical system in which the measurement of an observable
Q with a continuous spectrum [say spectrum Q = ( - 00, + 00)] gives
exactly one particular value, because such a state can be neither measured
nor prepared. It only makes sense to ask for a state of a physical system in
which the measurement of the observable Q gives values x lying in some
particular interval Xo - E < X < Xo + E around a mathematically defined
point xo. (Physically neither an exact point nor an exact interval, i.e., an
interval with precisely determined endpoints, can be defined.) Thus for
physical reasons, the analogue to (8.1a) for an observable Q with continuous
spectrum is
W= J dx' dx" h(x', xo)J.(x", xo)lx'><x"l, (8.5)
where hex, X o) is a function that has its main support in the neighborhood
(xo - t, Xo + E) of the point Xo and describes in a more or less direct way
the resolution of the measurement apparatus by which the state was prepared.
For the state W given by (8ola), the probability that a measurement of B
will yield the value b is
Tr(Ab W) = (b I WI b) = (b Ibo)(b o Ib) = 6bbo 6bob = 6bbo • (8.6)
For the state W given by (8.5), the likelihood that a measurement of Q will
yield the value x is described not by a probability but by the probability
distribution (probability density)
<xl Wlx> = J dx' dx" hex'; xo)i:(x"; xo)<xlx'><x"lx>
= J dx' dx" h(X'; xo)i:(x"; xo)6(x - x')6(x" - x)
= hex; xo)J:(x; Xo) = Ih(X; xoW = F,(x; x o)· (8.7)

Here we have used (7.19) [or (J.4.7c p ) if dx is symbolic for du(x)].
Symbolically we may write, in analogy to (8.6), <x x>
I W I = Tr(Ax W)
<x
with Ax = Ix> I; however, one has to keep in mind that Ax is not a projec-
tion operator onto a subspace of the space of physical states. The physical
question, i.e., the question that has an answer that can be compared with an
experimental determination, is: What is the probability of obtaining a value
x in the interval between Xl and X 2 when a measurement of the observable
Q is made on the state W? Here the interval (Xl' x 2 ) is determined by the
resolution of the apparatus with which the measurement is made, and again
cannot be infinitesimally small and exactly determined.
u.s Postulates U and III for Observables with Continuous Spectra 97
The operator which represents finding the position in the interval Xl <
X < X2 is the projection operator
(8.8)
the probability of finding the position anywhere in Xl < X < X2 when the
system is in the state W of (8.5) is then
Tr(A(x l , x 2 )W) = f dx' (x'IA(x l , x 2 )Wlx') = f XI

2
dx (xl Wlx)
(8.9)
The function F.(x; x o) = If.(x; xoW is determined by the experimental

situation. It describes the resolution of the apparatus by which the state W
is prepared. The better the resolution, the smaller is f, and in the idealized
but unphysical limit f --+ 0,
(8.10)
The limit (8.10) would correspond to an apparatus with infinitely sharp
resolution, which cannot exist in reality. Thus the state W in (8.5) can be
only as "pure" or as sharp as the experimental resolution allows, and that
can never be "ideally sharp" as described by a b-function.
In order that the state W given in (8.5) be normalized, f.(x, x o) must fulfill
Tr W = t+: dx (xl Wlx) = t+: dx If.(x, xoW = t+: dx F.(x, x o) = l.
(8.11)
Examples for F.(x) = F,(x, 0) are (function sequences of b-type):
1 f
Fix) = - 2 2 (Figure 8.1), (8.12)
n X +f
F (x) = _1_ ~ exp (- ~2 X2)
, fo f f2
(Figure 8.2), (8.13)
o if x < - f/2, }
F,(x) = {01/f if -f/2 < x < +f/2, (Figure 8.3), (8.14)
if + f/2 < x.
°
X+f
if x ::s; -f,
---;r if -f ::s; x ::s; 0,

Fix) = (Figure 8.4). (8.15)
-x + f
ifO::S;x::S;f,
f2
° if + f ::s; x,
i
II "
I
: III
I ~ t == O. 1
I \
I •
, I
, I
, I
I ,
- ----~ 0 -- - -- ---
~ ....
x
..
Figure 8.1 Graph of F£(x) = (l/rr)£/(x 2 + (2) for two values of £.
Figure 8.2 Graph of F£(x) = (l/~)(1/£) exp( -tx 2/(2).
Figure 8.3 Graph of F£(x) in Equation (8.14).
Figure 8.4 Graph of F( (x) in Equation (8.15).

11.8 Postulates II and III for Observables with Continuous Spectra 99
All these functions have their main support in the interval - f < X < +f
and have the property
F.(x) --+ J(x) as f --+ O.
Thus we have arrived at the basic assumption IlIa for continuous spectrum:
The statistical operator W given by (8.5) with a distribution function F.(x)
as narrow as experimentally possible is the continuous-spectrum analog of
a pure eigenstate.
We shall call such a state an "almost sharp" state and an "almost eigen-
state" of the observable with continuous spectrum.
An arbitrary state, in whose preparation no attempt at sharpness may have
been made, is described by (8.4). F(x', x") is connected with the probability
distribution function of the observable Q by
Tr(Ax W) = <xl Wlx> = f dx' dx" F(x', x")<xlx'><x"lx> = F(x, x). (8.16)
This is the appropriate place to mention the interrelationship

between the physical measurability and preparability of a state
and the mathematical description of quantum physics in a rigged
Hilbert space II> c :1t c II> x, though we cannot give a detailed
description here. ls
The resolution function F.ex; x o) or f.(x; x o) that corresponds to
a realistic experimental apparatus must be a well-behaved func-
tion; e.g., it can be something like that given in (8.12)-(8.15) with
a small but finite f for the description of a state as "sharp" as the
preparation allows. One can make this precise by requiring that
f.(x; x o) be an element of the Schwartz space, which is a realization
of 11>. More appropriate still may be to require that Fix; x o) have
compact support. Those f.(x; xo)'s that are either distributions or
elements of the Hilbert space (i.e., Lebesgue square-integrable
functions that are determined only up to their values on a set of
measure zero) cannot describe the resolution of a realistic apparatus
(e.g., the detector efficiency).
If f.(x, x o) is an element of the Schwartz space, then the state one
obtains by (8.5) or (8.4) is in the space 11>. More precisely, W is a
positive definite Hilbert-space-bounded trace-class operator with
W:1tc <1>. If W is a pure state, W= If><fl, thenf(x) = <xlf>
given by (8.3) is an element of the Schwartz space, and I f> is an
element of 11>. Thus the requirement that the space of physical states
consist of only those states that can be prepared in a realistic experi-
ment means that in a precise formulation of the preceding basic
assumptions the space of physical states should not be the Hilbert
space :1t, but the nuclear topological space II> of the rigged Hilbert
18 An elementary discussion of this subject is given in A. Bohm, The Rigged Hilbert Space
and Quantum Mechanics, Springer Lecture Notes in Physics, vol. 78, 1978.
space <l> c Yf c <l> x. Though the elements of <l> x which are not in
<l> do not represent physical states, <l> x is of immense value, as it
contains the generalized eigenvectors of self-adjoint operators with
continuous spectra, which are not elements of Yf; cf. Sections 1.4
and 1.7.
The Hilbert space Yf is infinite-dimensional, and it is infinite-
dimensional in a very particular sense; namely, it is complete with
respect to a particular topology, that is, with respect to a particular
meaning of convergence of infinite sequences. Since an infinite
number of states can never be prepared, physical measurements
cannot tell us anything about infinite sequences, but can at most
give us information about arbitrarily large but finite sequences.
Therefore, physics cannot give us sufficient information to tell how
to take the limit to infinity, i.e., how to choose the topology. The
choice of the topology will be a mathematical generalization, but
then one should choose the topology such that it is most convenient.
With the choice of <l> as the space of physical states, one does not
only obtain a description which is closer to reality for the above-
mentioned reasons connected with measurement and preparation
of physical states, but one also obtains a tremendous simplification
of the mathematical description. Even the simplest algebra of
operators that appears in quantum mechanics, the one fulfilling the
relation (2.1a), cannot be represented by Yf-continuous operators.
Consequently there are vectors that lie outside the domain of the
definition of an observable that is represented by an unbounded
operator in Yf, e.g., the vector that would correspond to a state of
infinite energy. In contrast to this, it seems possible that for all
physical systems the observables can be represented by <l>-contin-
uous operators. Thus in the rigged-Hilbert-space description the
mathematical image of a physical system is an algebra of continuous
operators, domain questions do not arise and all the mathematical
manipulations that physicists perform are rigorously justified.
After having considered the continuous-spectrum analogue (8.5) of (4.31)

in the axiom IlIa we shall now consider the basic assumption IIlb for a
continuous spectrum. Let the physical system be in a state W; this state
may have been prepared by measurements of an observable with a con-
tinuous spectrum or with a discrete spectrum, or may even be completely
unspecified. If a measurement of an observable Q with continuous spectrum
- 00 < x < + 00 is made on the state Wand is used to select out a sub-
ensemble with values x of Q lying in the range Xl < X < X 2 , then the (un-
normalized) state W' that describes this subensemble is
W' = A(x l , x 2 )WA(x 1 , x 2 )
= IX2
Xt
dx'
JX2
Xl
dx" Ix')<x'i Wlx")<x" I. (8.17)
This is the continuous spectrum analog of (4.52). Note that for Xl -. - 00,
x 2 -. - 00 we have W' = W, which says that no measurement has been
performed. Again, Xl < X < x 2 must be a finite-size interval; it cannot be
infinitesimally small, due to the finite resolution and finite extension of the
measuring apparatus.
To show that (8.17) represents a state where measurement of Q must
yield a value X in the interval Xl < X < x 2 , we compute the probability
density for getting the value x, which according to (8.17) is given by
1,
= --
Tr W
f f
X
Xl
2
dx'
X
Xl
2
dx" <xlx')<x'i Wlx")<x"lx)
{o if X < Xl or X > x 2 '

(8.18)
= (Tr W')-l<xIWlx) if Xl < X < x 2 •
An actual position measurement is over some range of values in a small
interval Xo - £ < X < Xo + f. The probability for measuring Q to be in
such an interval is
<A(xo - £, Xo + f»~'
o
{
-
_ 1 Xo+'
~f dx<xIWlx)
r W Xo-'
(8.19)
Thus W' represents a state where the position is in the interval (Xl' x 2 ).
11.9 Position and Momentum Measurements-

Particles and Waves
The material of Section 11.8 has provided us with a formalism in which we

can discuss problems connected with the measurement of position and
momentum in more detail. Let us return to the classical-mechanical model
of the harmonic oscillator, which consisted of two mass points vibrating
around an equilibrium point. For simplicity we assume that m 2 ~ ml ,
so we can consider m 2 to be at rest. The position X and momentum pare
then essentially the position and momentum of mass point l.
How can a measurement of the position be made on the quantum-
mechanical harmonic oscillator? We have considered the diatomic molecule
as a realization of this system. The question is, then, how do we measure the
position, i.e., the separation between the two atoms of the CO molecules in
the collision chamber? There seems to be no way one can think of to set up
a measurement apparatus that would do so. The same argument also applies
to a momentum measurement on this ensemble. Thus the above discussion
in Section 11.7 of the probabilities w(x) and w(p) is not of much practical value
if one restricts oneself to the quantum-mechanical harmonic-oscillator
system. With the formalism developed in Section 11.8 we can give a theoretical
explanation of this situation.
According to (8.17), after a measurement of the position operator Q,
the physical system is in an almost eigenstate of the position operator; i.e.,
all molecules in the ensemble have a position localized in the interval
(x - f, X + f). For our classical model of the CO molecule as two oscillating
mass points (with m 2 ~ m1 for simplicity), the above statement means that
m1 is located at a distance x from m2 that lies in the interval (x - f, X + f).
However, a classical system for which m1 is located in a restricted domain
of space is not an oscillator. It is a localized mass point, which we call a
particle. A measurement of the position operator Qon the harmonic oscillator
would destroy the oscillator and result in a physical system that has as
its classical image a particle. Thus the fact that it is impossible in practice
to set up an experiment that would measure the observable Q on the en-
semble of CO molecules in the collision chamber has a theoretical description
in the basic assumption Illb as described by (8.17).
The above arguments also exemplify the fact that for a particular physical
system, such as the vibrating CO molecule, there are different kinds of
observables. Of one kind (e.g., the energy operator of the vibrating CO
molecule) one can easily prepare eigenstates of the particular physical system.
Of the other kind (e.g., the relative-position operator of the vibrating CO
molecule) the preparation of eigenstates will destroy the system.
The measurement of this second kind of observable can only be described
if one enlarges the physical system. Thus in our example we need not restrict
ourselves to CO molecules, but can consider the larger physical system that
consists of CO molecules, 0 atoms, and C atoms. Then the CO molecules
are just particular states of this larger physical system. Other states are the
states in which the C and 0 atoms are not bound. In this larger system, the
measurement of the position (the distance between C and 0 atoms) is always
possible at least in principle.
For the almost eigenstate of the position operator Q, we have
w(x) = II dx' dx" ~(x' - x)~x" - x)lx')<x"l, (9.1x)
where Qlx) = xix) and I~(x' - x)1 2 has the property of the functions
(8.12)-(8.15). One can define the almost eigenstate of the momentum
operator P,
W(p) = I:oo dp' Loooo dp" j.(p' - p)J.(p" - p)lp')<p"l, (9.1p)

where Pip) = pip) and I!,(p' - pW is again a function of the kind given
in (8.12)-(8.15). Q and P here are related by (2.1a). In such a state the mea-
surement of the momentum P gives a value pin the interval p - ( < P< p + (.
By the same arguments as above, such a state cannot be realized by the
oscillator; therefore we do not want to require (2.1b), and shall just require
that Q and P be related by (2.1a) and that as a consequence 19 of this <pix)
is given by (7.50). A possible relation between the energy operator Hand
the P and Q that facilitates the preparation of almost momentum eigenstates
is H = p2/2m, where m is a system constant, the mass. A quantum physical
system obeying this relation is called a free nonrelativistic elementary
particle.
A quantum-mechanical system with well-defined momentum has as its
classical image a wave.
Let us discuss this last point a bit more. The characteristic or defining
property of a particle is that it is localized in space, i.e., it can be assigned a
definite position. In the same way, the characteristic or defining property of
a wave, in particular a plane wave, is that it has infinite periodic spatial
extension. The mathematical description of a plane wave is
(9.2)
so the periodic spatial extension is described by
u(x) = Ae ikx , (9.3)
where A is the amplitude and k is the wave number. The fundamental period
or wavelength is A. = 2nlk. If the wave is not a plane wave, i.e., does not have
a well-defined wave number or wavelength but has wave numbers k' inside
a certain interval (k - 11k, k + 11k), then one has a wave packet described by
¢(x) = f+OO
_ 00 dk' A(k')eik'X, (9.4)
where A(k') is nonzero only in the interval 2 11k about the value k.
To see that a quantum-mechanical system with well-defined momentum
has as its classical image a wave, let us calculate the spatial distribution of a
state with well-defined momentum. Let us first do this simple-mindedly,
working with "states" having exact momentum Ip) and "states" having
exact position Ix) (which are "described" by W(p) = dp Ip) <pi and W(x) =
dx Ix)<xl, respectively), though we should keep in mind that such "states"
are not physical states.
For a pure state AIjJ with "well-behaved" state vector t/I, the spatial distri-
bution is given by the wave function t/I(x) = <x It/I); according to (5.25) the
probability density for measuring the value x in a position measurement of
the state t/I is wljJ(x) = l<xlt/l)1 2 . Now let t/I be the "not well-behaved" and
19 As remarked in [] at the end of Section 11.7, (2.la) is not sufficient to derive (7.50). One
has to make an additional assumption, e.g., that p2 + Q2 is essentially self-adjoint.
therefore unphysical "state" vector Ip). The spatial distribution is then

given by (xlp), which is, according to (7.51),
(xlp) = -1- e,xPI

"h
. (9.5)
foh
If we compare this with the spatial distribution of the plane wave given by
(9.3), we see that an "exact momentum eigenstate" of a quantum-mechanical
system has the same spatial distribution as the classical system called a plane
wave, with the wave number and wavelength given by the de Broglie relation:
1 2nh h
k =-p and A= - =- (9.6a)
h p P
This is the justification for calling an ensemble of quantum-mechanical
systems with definite momentum p a wave with wavelength A = hlp.
Let us now return to the physical case where the state is given by (9.1p)
and is an almost momentum eigenstate with as narrow a momentum dis-
tribution I h(p' - pW as possible. The probability density for obtaining
the value x in a measurement of the observable Q in the state W(p) is given
according to (8.7) by
(xIW(p)lx) = t+: t+: dP' dP" h(p' - p)J,(p" - p)(xlp')(p"lx)
=
f _+OO dp' h(p'
00 - p)~
ixp'/h f+oo
_ 00 dp" J,(p" - p)
e-ixp"/h
foh (9.7)
In order to compare this with the wave function for a (classical) wave packet
f
(9.4), we define
+ 00 eixp'/h
¢(x) == _ 00 dp' h(p' - p) foh
= f+ 00 dk' h h(k' - k) eixk', (9.8)
-00 J2iih
If
where p' = hk'. With
A(k') = h(k' - k),
¢(x) indeed describes a wave packet (9.4), and the intensity of this wave
gives, according to
(xl W(p)lx) = 1¢(x)1 2 , (9.9)
the probability density for the value x.
Thus a quantum-mechanical system in a state with well-defined momen-
tum (in a narrow momentum interval) has as its classical image a wave
packet with wavelength A in a narrow interval:
2nh > A > 2nh . (9.10)

P-f. P+f.
11.9 Position and Momentum Measurements-Particles and Waves 105
The intensity of this wave corresponds to the probability density for the
measurement of the position. </J(x) describes the probability and is called
the probability amplitude.
We now understand why ljJ(x) = <x 11jJ) has been called a wave function:
Let W", be a pure state, i.e., W", be a projector onto the space spanned by 1jJ.
Then the probability density for obtaining the value x in a measurement of
the position is
W",(x) = <xl W",lx) = <xlljJ)<ljJlx)

= ljJ(x)lJj(x) = IIjJ(x)12. (9.11)
On the other hand, we can calculate <x I W", I x) by inserting a complete

system of generalized states
I = f dp' Ip') <p'l,
to obtain
<x I W",lx) = f f
dp' dp" <x I pl)<p'IIjJ)<1jJ Ip")<p"lx)
= f fi; f
dp' ljJ(p') dp" e~h ljJ(p"). (9.12)
If we compare (9.11) and (9.12), we see that the wave function, whose physical
interpretation was that its absolute value squared is the probability distri-
bution, is given by
ljJ(x) = f fi;
dp' ljJ(p') (9.13)
with ljJ(p') = <p'IIjJ). Thus the wave function ljJ(x) may be considered as a
superposition of plane waves, ljJ(x) comes closer to being a plane wave as
the transition coefficient <p'lljJ) between the pure state IjJ and the generalized
momentum eigenvector Ip') comes closer to being aD-function b(p' - p).
The wave character of the quantum system can be experimentally dis-
played by diffraction experiments. Let us consider a system that has been
prepared in an almost eigenstate of momentum, i.e., a monochromatic beam
of quantum particles. Such a monochromatic beam is easy to prepare if the
particles are charged, but harder to prepare if they are neutral (like atoms
or molecules). Let us therefore consider first a monochromatic beam of
electrons, such as that prepared for the electron-loss experiment that we
considered before. The electron beam is produced by a filament, then ac-
celerated by an electric potential, and then passed through crossed electric
and magnetic fields of very well-defined field strengths. Only electrons with
very well-defined momenta are not deflected. If f.(p' - p) describes the
momentum spread of this beam, then the spatial distribution is given by

(9.8):
cjJ(x) = fdk' h f.(k' - k) . ,

.jhl e,xk, (9.14)
which becomes closer to being a plane wave,
as the momentum spread about the value hk becomes narrower. This ap-
proximate plane wave may be diffracted by a grid, and one obtains a typical
diffraction pattern.
Related to Equation (9.6a) by special relativity is another equation:
1
W =1zE, (9.6b)
where w is the angular frequency of light, w = 2nv. The connection between

(9.6a) and the relation (9.6b) is best obtained from the dispersion law for
our waves, W = f(k), and the relation between energy and momentum,
E = f(p). Thus, for the system "electromagnetic wave-photon" the dis-
persion law is w = ck (v = cjA), and the relation between energy and mo-
mentum is E = cp [E2 - (Cp)2 = (mc 2)2 with rest mass m = 0]. From these
classical relations (the first for a wave, the second for a massless relativistic
particle) it follows immediately that (9.6b) is a consequence of (9.6a) and
vice versa. For the system" electron de Broglie wave-nonrelativistic electron"
the dispersion law is w = k2 j2(mjh) and the relation between energy and
momentum is E = p2j2m, which again shows that (9.6a) and (9.6b) are
consequences of each other.
The two physical systems "electromagnetic wave-photon" and

"electron de Broglie wave-electron" are two examples of physical
systems called elementary particles. 20 The historical development of
the ideas about these two systems was, however, quite distinct: To
begin with, all known experimental facts about light could be
explained by a particle theory (Newton, 1663) as well as a wave
theory (c. Huygens, 1678). The same domain of physical knowledge
could be described by two different pictures. Then, diffraction of
light was discovered (Young, 1801), which could only be explained
by the wave picture, not the particle picture. The cathode rays, on the
other hand, though considered by some to be, like light, due to a
process in the ether, were soon proven to behave like negatively
charged particles in an electric and magnetic field (1. 1. Thomson,
20 It would have been better to coin a new name for these physical systems (e.g., "quanta ")
to avoid any reference to Newtonian particles.
H.9 Position and Momentum Measurements-Particles and Waves 107
1897; 1. Perrin, 1897). Thus light was a wave and electrons were
particles. Had the photoelectric effect (P. Lenard, 1902) been
discovered before the diffraction phenomena of light, and had the
electron diffraction (Davisson and Germer, G. P. Thomson, 1927)
been the first effect observed with the cathode rays, the situation
might have been just the reverse.
As light was considered a wave, one had the dispersion law
w = ck, or w = c/A, and Einstein (1905) showed that in order to
explain the photoelectric effect one has to associate with light,
besides the energy given by Planck's (1900) formula E = hw, the
momentum p = E/c.
Electrons, on the other hand, were considered particles, and one
had the energy-momentum relation E = p2/2m. But de Broglie
(1924) postulated that they are waves, with a wave number given
by (9.6a). Then he derived the dispersion law for these waves from
the Planck formula (9.6b) to be w = k 2 /2(m/h).
The two relations
E = hw, p= hk (9.6)
are justly considered the gate to quantum theory. We derived at least the
second equation (9.6a) from the defining relation (2.1a). But historically
they were the starting point of the long conjecturing process that led to
(2.1a) and ultimately to quantum mechanics in general. The relations (9.6)
connect two different classical pictures (often called dual).
The one picture is a particle, i.e., a localizable mass point. Classically,
such a particle can be given definite values of energy and momentum (e.g.,
the momentum and the energy measured by an apparatus that moves with
a velocity - v relative to such a localized mass point are p = mv and E =
p2/2m).
The other picture is a wave, i.e., a more or less periodic disturbance of
infinite spatial extension. A wave has a frequency wand a wave vector k.
In quantum mechanics the physical system (quantum, elementary particle),
of which we should always imagine an ensemble, can occur in different
states. For instance, it can be in an almost momentum eigenstate described
by (9.1p). We may then ask about the position of such a state, e.g., we may
want to know the probability of finding the position anywhere in an interval
x - 11/2 < x' < x + 11/2. This is according to the basic assumption II,
expressed with the projection operator of (8.8) as:
((
Tr A x -
11 11) W(p) ) =
2' x + 2
fX
+/;./2
x-/;./2 dx' <x'i W(p)lx').
For the sake of definiteness we may assume that the almost momentum
eigenstate is given by (9.1p) with F.(p' - p) given by (9.14) or Figure 8.3,
so that J.(p' - p) = 1/~ in the interval between p - f/2 and p + f/2. The
probability density of the position operator is then obtained using (9.7) as
J P +£/2 1 eixp'lh f P +£/2 1 e-ixp"lh

<xIW(p)lx) = dp'--- dp"----
p-£/2 ~ jbJz p-£/2 ~ jbJz
1
= ~ f for fx/Ii ~ 1. (9.15)
2nH
This means that the probability density of the position operator is indepen-
dent of the position. Thus the probability of finding for the physical system
a value in the interval of length Ll around x is
f X +tJ./ 2
x-tJ./2
1
dx' <x'i W(p)lx') = 2h fLl,
n
(9.16)
i.e., the system is located with the same small probability anywhere in space.
In the limiting case of exact momentum when the momentum spread
f --+ 0, the probability of finding the physical system in an interval Ll any-
where would be zero. However, this limiting case is unphysical, because
there is no apparatus that could prepare such a state. For example, an almost
momentum eigenstate of electrons, a monochromatic electron beam, can
be prepared by the apparatus used in the energy-loss experiment of Section
11.4. The electrons produced by a filament are accelerated by an electrical
potential. Even if this potential is very well defined, the electrons will have
still the same momentum spread with which they were emitted by the fila-
ment. This electron beam then passes through a monochromator (an electric
or magnetic field, or in some cases crossed electric and magnetic fields), and
only the electrons with a definite velocity or momentum will arrive at the
position of the slit. But again, even if the field is very well defined, this slit
has a finite opening, so that electrons with slightly different momenta can
pass through. In general, because of the necessarily finite size of the macro-
scopic measurement apparatus, the preparation of a monochromatic beam
with zero spread is impossible.
As we have seen above, for an almost momentum eigenstate every position
has the same probability, i.e., the position is undetermined. The physical
system is in a wave state with a rather well-defined wave number and wave-
length; thus it appears as a wave.
Thus, according to the formulation of quantum mechanics based upon the
basic assumptions I, II, and III, a quantum physical system in a momentum
eigenstate behaves like a wave-it "is" a wave-until a measurement is
performed that alters the state (e.g., a position measurement). As long as
such a measurement is not performed, the theoretical predictions are exactly
the same as in a classical wave theory.
We can go through the same arguments starting with an almost position
eigenstate W(x) described by (9.1x). The probability density of the momen-
tum operator is then obtained as
1
<pi W(x)lp) = 2nh e, (9.17)
H.9 Position and Momentum Measurements-Particles and Waves 109
and the probability of finding the momentum in an interval between p - (./2

and p + (./2 is
fP +(/2 1
dp' <p'l W(x)lp') = 2h ~L (9.18)
p-£/2 7C
As this probability does not depend upon the central value p, any value
of p has the same probability. For an almost position eigenstate the momen-
tum is undetermined. Such an almost position eigenstate we have called
a particle, but we see now that, in distinction to a classical (Newtonian)
particle, the quantum physical system in a particle state cannot have a
definite value of momentum.
The property described by
W(p) = A(P - ~, P+ ~). (9.19)
i.e., that the physical system is in a state with well-defined momentum, and
the property
W(x) = A(x - ~, x + ~). (9.20)
i.e., that the physical system is in a state with well-defined position, are called
complementary to each other. The realization of one property precludes any
prediction for any complementary property. If a physical system is in the
state (9.19) of well-defined momentum then nothing can be said about its
position; any position x is equally probable or improbable as expressed by
(9.15). If a quantum physical system is prepared to occur with the classical
attributes of a wave then it cannot bear the classical attributes of a particle
and vice versa.
If the properties Xand A are complementary then the properties 1 - A
and A, X and 1 - A, and 1 - X and 1 - A are also complementary. 1 - A
is the negation of the property A: If A is the property that the quantum
physical system is in the space domain between x - ~/2 and x + ~/2 then
1 - A is the property that the system is outside this domain. However, in
distinction to the classical case, the quantum physical system need not have
the property 1 - A if it does not have the property A; e.g., it could be in the
state X, i.e., it could occur as a wave and have no position at all or, instead
of X, it could as well have any other property incompatible with A and 1 - A.
There are thus an infinite number of possibilities for a quantum system to
have neither the property A nor its negation 1 - A, whereas a classical system
would always have either the property A or its negation 1 - A.
The constant Ii was introduced in the defining relation (2.1a) without
much discussion. With the consequences of (2.1a) derived so far we can
already understand the meaning of n.
nis a conversion factor for the values of one and the same physical quantity
in different units. Momentum and energy can be measured in the conven-
tional ways. One prepares an electron beam with sharp momentum by, e.g.,
using the apparatus of Figure 4.2(a): Electrons, first treated as particles, are
emitted by the filament and accelerated by the potential difference of an
electric field to the velocity v which can be calculated from v2 = 2(e/m)V.
With V measured in volts and with the usual value
~ = 1.7588.1011 Coulomb = 1.7588.10 5 cm 2

m kg ~dv
one obtains v in cm/sec or p = mv in erg sec/cm.

This electron beam is then treated as a wave. It is diffracted by a grid
(grids with the right grid constants of approximately 10 - 8 cm = 1 A are
crystals) and its wave number k is measured. Using our present knowledge
that particles and waves are just two different appearances (of many possible
appearances) of the quantum physical system "electron" for which the
particle property v = dx/dt = velocity of the particle is related to the wave
property A. = 2n/k = wavelength of the wave by the relation (9.6a):
h k h 2n
v = - =-- (9.21)
m m A. '
one can determine h/m from the value of v measured with the electric field
and from the value of A. measured with the grid. One need not restrict oneself
to this particular experiment only, but one can do many more experiments
with electrons measuring independently v from the particle property of the
electron and A. from its wave property. The value of h/m that one obtains
using (9.19) will be the same. Thus h/m is a constant as long as one uses only
electrons.
Now one can repeat the same procedure with other physical systems, e.g.,
protons, and one can again determine the value of h/m from Equation (9.21).
One finds that this leads to a different value of h/m than in the experiments
with electrons. Thus the conversion factor h/m between k and v depends upon
the nature of the micro system.
But m can be measured independently-and it is conventional to use for
it the independent unit g-by first measuring elm in deflection experiments
with cathode rays and proton beams and then by measuring e in a Millikan
experiment. In this way one finds that h, the conversion factor between k in
cm - 1 and p = mv in g cm/sec,
mv = hk, (9.6a')
has the same value for all physical systems, namely21
cm
h = 1.0546.10- 28 g - cm
sec
= 1.0546.10- 28 erg sec = 6.58218.10- 16 eV sec. (9.22)
The conversion factor h between wave number k and momentum p has
nothing to do with the structure of the microphysical system. It is not a system
21 1 eV = 1.6 x 10- 12 erg = 1.6 x 10- 12 dyn em; 1 dyn = 19 em/see. 2

11.9 Position and Momentum Measurements-Particles and Waves 111
constant like J.l and ill in the microphysical system of Section 11.2, or like
(h/m); it is a universal constant.
Why is the conversion factor between k and va system-dependent constant
and the conversion factor between k and p an universal constant of nature?
The reason for this lies in the structure of the symmetry group of space-time,
the Galilei group (similar arguments also apply to the symmetry group of
relativistic space-time, the Poincare group). The observables G/h = mQi/h
and P /h-obeying the commutation relations [G/h, P/h] = ilJi/m/h)
(i, j = 1, 2, 3), from which the three-dimensional version of (2.la) follows-
are not just observables of a particular micro system or of various micro-
systems but they represent the generators of the Galilei transformations:
(l + ixiP/h) represents the displacement by an infinitesimal vector Xi (in
cm), and (l + iviGdh) represents an infinitesimal proper Galilei transforma-
tion (to a frame moving with velocity Vi in cm/sec). Thus Pi/h and mQ/h
(but not Q/h) have a universal meaning which comes from the structure of
the symmetry group in nature and not from the structure of a particular
physical system. m/h is an invariant of the group of Galilei transformations,
whose value characterizes the representation of the (extended) Galilei group
and, therewith, the physical system that this representation describes.
The conversion factor h appears in the expression for the Galilei trans-
formation because Xi and Pi have in the past been chosen to have independent
units: Xi is measured in cm and Pi is measured in erg sec/cm, a unit which is
derived from the fundamental unit g for the mass
( 1 erg -sec = 1 dyn cm -sec = 1 g -cm sec)

2 cm - .
cm cm sec cm
But this choice is no longer necessary now that one knows quantum mechan-
ics and the fundamental physical meaning of the representations of the
Galilei group. One can decide to choose the units such that h = 1 because all
one needs is that xip/h and vimQ/h be dimensionless quantities. Taking for
Xi the conventional units cm, the choice h = 1 means that Pi has the units
cm - 1 and that the units of the mass are then no longer the independent
units g but are the derived units sec/cm 2 • The value of h in the conventional
c.g.s. system, which we denote by {h} and which according to (9.22) is given
by {h} = 1.0546.10- 28 , is therefore nothing other than the conversion factor
between g and sec/cm 2 • Thus choosing the conversion factor h = 1, only
cm and sec remain as fundamental units.
For relativistic space-time one has an additional conversion factor
between a distance X measured in cm and the time t (in sec) that light takes
to traverse this distance, t = (l/e)x, where e is the velocity of light. Choosing
e = 1 for this conversion factor eliminates sec as an independent unit and
1 sec = {l/e} cm, where {e} = 2.9979.10 10 denotes the value of e in the
c.g.s. system. Then cm remains as the only fundamental unit.
Instead of cm - lone usually takes in relativistic particle physics the units
of 1 eV or 1 MeV = 106 eV for energy, mass and momentum. The distance
is then measured in MeV- 1. The value {he} of he in the c.g.s. system of units
is then the conversion factor between the units of energy, mass, and momen-
tum in Me V and in cm - 1 :
I cm- 1 = {lie} MeV = 1.97328.10- 11 MeV.
Thus we have seen that Ii, like any constant of nature, is just a conversion
factor for the values of one and the same physical quantity in different units.
Before a constant of nature is discovered, the values of the same quantity
in different units (e.g., the momentum p and the wave vector k) are thought
to be different physical quantities. After the constant has been discovered
and the different physical quantities have been related by a new theory, one
can measure these quantities in the same units and eliminate in this way one
of the previously independent units.
Problems
1. Show that the operators a and at, defined by Equations (3.1), are the adjoints of
each other.
2. Show that the collection of all linear operators on a (complex) linear space R
constitutes an associative algebra with unit element under the operations of scalar
multiplication of an operator, addition of operators, and multiplication (composition)
of operators.
3. For the quantum-mechanical harmonic oscillator, calculate the diagonal matrix
elements of the energy operator H between the states
ct." = real,
where cp" is the eigenvector belonging to the eigenvalue n of the operator N = ata,
according to Equation (3.20).
4. Let cp" be the vectors of Equation (3.19). Define for each complex number z the
coherent state vector 22
zrl
I ----= CPn'
00
Iz) == e-IzI"2
FO yin!
(a) Show that the coherent state Iz) is an eigenstate of the operator a with eigen-
value z.
(b) Show that the scalar product of cP" with Iz) has the property
1<<p"lzW
Iz1 2 "
= _~e-lzI2
n!
(c) Show that the scalar product <:: I::') of two coherent states I::) and Iz') is given by
<zl::') = exp(-tl:12 + 22' - ~lz'12).
(d) Calculate the expectation value of the operators H, Q, and P in the coherent
state Iz), i.e., calculate <zIHlz), <zIQlz), and <zIPlz).
(e) Show that for the pure coherent state W = Iz)<zl the equality holds in (6.3):
I1P I1Q = l
22 For the applications of coherent states consult Sargent Scully Lamb (1974), Ch. XV.
Problems 113
5. Figure PS.l(a) shows a schematic diagram of a slightly different experimental

setup than the one of Figure 4.2(a). Results of an energy-loss experiment performed on
N2 with this apparatus are shown in Figure PS.l(b). The bump at v = Oin Figure PS.l(b)
is approximately 30 times larger than the bumps at v = 1,2,3, ... ; this is in contrast
to the situation depicted in Figure 4.2(b), where the v = 0 bump is approximately three
times larger than the succeeding bumps. Explain the difference. Determine from Figure
o
~
;;
v
o
Swttp VolYejo, ""II,
(b)
Figure PS.l (a) Schematic diagram of the double electrostatic analyzer. Electrons are
emitted by the filament , deflected by the cylindrical grids (4A and 4B) with radii of 1.0
and 1.5 cm, respectively, injected into the collision chamber 5, and analyzed by sweeping
the voltage between electrodes 6 and 3. S p S2' S3' S4' S 5 are shields to collect unwanted
electrons; 4C and 7C are top and bottom grids. Typical operating voltages between the
electrodes indicated: (4A - 4B) = 1.2 V; (7A- B) = 1.2 V; (Fil- 3) = 1.4 V; (F- S t ) =
20 V; (F - S2) = (F - S 3) = 20 V. The electron collector is grounded. All slits are
0.5 x 4 mm. (b) Energy spectrum of forward-scattered electrons at an incident electron
energy of 2.6 eV. [From Schulz, G. J: Physical Review 125,229 (1962), with permission.]
PS.l(b) the statistical operator W that describes the state of the ensemble ofN 2 molecules
of the experiment of Figure PS.l(a).
6. Suppose we have a quantum-mechanical oscillator.
(a) Let the state of the system be given by (4.17). Show that (4.21) follows from (4.20).
(b) Show that W = WI(Tr W) is the normalized statistical operator, where Wis
given by (4.22).
(c) Let A be a projection operator. Show that Tr A = dim(A.n").
7. Calculate the expectation values of the operators Q, Q2, P, and p 2 when the quan-
tum-mechanical oscillator is in the state W = An' where An is the projection operator
onto the energy eigenspace corresponding to the eigenvalue En =1fw(n + t).
8. Let A be the projector onto the one-dimensional subspace spanned by tjJ (ljtjJll = 1).
Show that the dispersion disp(A) H of H when the system is in the state W = A satisfies
disp(A) H z O.
9. Calculate disp(w) H for the state of the CO molecules in the collision chamber of
the energy-loss experiment described in Section 11.4. Show that disp(w) H z O.
10. Let W be given by Equation (5.9). Show that disp(w) A = O.
11. Let W be given by (4.31). Calculate (A) and show that this is the result we would
expect if (A) is to represent the average value of A. Is W' as given by (4.51) normalized,
i.e., does Tr W' = 1 ?
12. Let Hand B be two observables with spectra En and bi' respectively. Let An and
A b , be the projection operators on the eigenspaces of H with eigenvalue En and of B
with eigenvalue bi' respectively. Let the quantum-mechanical system be in a state in
which an energy measurement has always resulted in the value En'
(a) What is the statistical operator W for this state?
(b) In this state W a measurement of the observable B is performed. What is the
statistical operator W' after the B measurement, without selection of subsystems?
(c) What is the probability of finding the value b j in a measurement of B on the
physical system in the state W? in the state W'? Compare these two probabilities.
(d) The observable H is now measured again. What is the probability of obtaining
En in this measurement of H? Compare this probability with the probability of
obtaining En in the original state W.
(e) Let A be a third observable with eigenvalues ai and projectors Au, onto the eigen-
spaces. Calculate the probabilities of obtaining the value ak in a measurement of
A when the system is in the state Wand when the system is in the state W'. When
are these probabilities equal?
13. Calculate the matrix elements (nlPlm) and <nIQlm) of the operators P and Q
between the energy eigenvectors ¢n = n) of the harmonic oscillator.
1
14. Give explicit expressions for the first six Hermite polynomials
Hb) (n = 0, 1,2,3,4,5)
and discuss the probability distribution 1(x 1n) 12 for the harmonic oscillator in the six
lowest energy eigenstates.
15. In the energy eigenstate W = A 1 corresponding to the energy eigen value E 1 = ~hw,
the probability w 1 (x) L'lx of measuring the position to be within an interval
Problems 115
(x - lu/2, x + I1x/2) about the value x is nonzero for x > a = J2Ed(l1o/). From
this it follows that the potential energy V may be larger than the total energy E 1 because
Point out what is wrong with this argument, and give a correct interpretation of the
situation.
16. Show that the operators p(S) = (h/i) d/dx and Q(S) = x, considered as operators
on the wave function, satisfy the Heisenberg commutation relation.
17. It was derived in Section 11.7 that, on the wave function, the momentum operator
p and the position operator Q are represented by the operators p(S) = (h/i) d/dx and
Q(S) = x. Which assumption besides the Heisenberg commutation relation has been
used in this derivation?

18. Let a quantum-mechanical particle with mass m be confined by impenetrable
walls to the region -a < x < +a (a one-dimensional rectangular well with infinitely
high walls). This means that the expectation value of the potential-energy operator V
between the generalized position eigenstates Ix) is
<V) = <xlVlx) = -c,

x (xix)
where c is a constant; and that the probability that a measurement of the position Q
will give the value x is nonzero only for -Q < X < +a. Denote by In) the eigenvectors
of the energy operator
p2
H=-+ V
2m '
and let An denote the projector onto the space spanned by In). Calculate the eigen-
values of H, and compute the expectation values of Q, Q2, P, and p2 when the system is
in an energy eigenstate An'
19. Show that for any linear operators A, B and for ..t E C, the trace as defined by
Equation (4.3) has the following properties:
(a) It is independent of the choice of basis.
(b) Tr(..tA) = ..t Tr A.
(c) Tr(A + B) = Tr A + Tr B.
(d) Tr(AB) = Tr(BA).
20. Let W(p) be an almost momentum eigenstate described by the statistical operator
W(p) = f +OO dp' f+oo dp" J:(p' - p)l,(p" - p)lp')<p"l

- co - C()
with
, 1 ~
J,(p - p) = I: ' .
ynP -P-l£
(Lorentzian momentum distribution). Calculate the probability density of the position

operator and the probability of finding the position in the interval Xo - I1x/2 < x <
Xo + I1x/2. Discuss the result.
21. Davisson and Germer scattered low-energy electrons from metal targets. For
45-eV electrons incident normally on a crystal face, compute the angle between the
incident beam and the scattering maximum if the metal is assumed to be of simple
cubic structure with a lattice constant of 3.52 A.
22. Has it been derived that electrons are waves with wavelength A = hip? If so, then
explain how this has been done. If not, explain what has been derived.
CHAPTER III
Energy Spectra of Some Molecules
Section 111.1 discusses how the energy levels of a quantum-mechanical

system emitting dipole radiation are observed. A derivation of the transition
probability is not given. In Section IIl2 the defining relations of angular
momentum are established. Section 111.3 derives the representations of the
algebra of angular momentum. In Section 111.4 the energy spectrum of a
rotator is derived and compared with the experimental spectrum of diatomic
molecules. Section I1l5 contains the basic assumption about the physical
combination of two nonidentical quantum-mechanical systems and the
application of this assumption to the description of vibrating and rotating
diatomic molecules.
111.1 Transitions Between Energy Levels of Vibrating

Molecules-The Limitations of the Oscillator Model
A quantum-mechanical system in a certain stationary state [e.g., the energy

eigenstate An of the diatomic molecule (oscillator)] will remain in that state
so long as it is not acted upon by outside forces. In practice, any quantum-
mechanical system is acted upon by weak external forces, such as external
electromagnetic fields or internal electromagnetic fields that arise from the
motion of charges within the system. Under the influence of such forces, the
state is liable to change. If the system has a discrete set of states (e.g., the
energy eigenstates of the oscillator), then a weak external disturbance does
not change these states (or, more precisely, it changes the energy levels by a
117
118 III Energy Spectra of Some Molecules
negligibly small amount), but the system may jump from one state to another.
The theory of such transitions, which can be developed as a consequence of
the basic assumptions of quantum mechanics, will be presented later in
Chapters XIV and XXI. For the moment we shall just give some semi-
classical arguments and state the result, which we shall use to obtain transi-
tion frequencies and selection rules.
We accept here as an empirical fact that under the influence of ever-
present external disturbances, the quantum system may perform transitions
from one energy eigenstate with energy En to another with energy Em, and
emit or absorb the energy difference
En - Em
as electromagnetic radiation in the form of a light quantum or photon of
frequency
(1.1)
If the electromagnetic field has the frequency Vnm , then the quantum-mech-
anical system can absorb a photon of this frequency and jump from a state
of energy En to a state of higher energy Em. On the other hand, if a quantum
system is in an excited state En (a state of higher energy than the ground
state), it can emit a photon of frequency Vnm and drop to a state of lower
energy Em.
Transitions between two states cannot occur under the influence of electro-
magnetic radiation if the matrix element of the total electric displacement
operator D of the system vanishes between these two states. 1 Also, the proba-
bility for such a transition, and thus the intensity of the emitted (or absorbed)
electromagnetic radiation, is proportional to the square of the modulus of
this matrix element.
To illustrate, let us return to our classical picture of the diatomic molecule.
If the molecule consists of unlike atoms (e.g., CO) then it has an electric dipole
moment, since the centers of the positive and negative charges do not coincide.
The dipole moment is the vector directed from the center of negative charges
to the center of positive charges and is given by
D = qd,
where q is the charge and d is the distance between the centers of the charges.
The permanent dipole moment Do ofthe molecule lies along the internuclear
aXIs.
If the interatomic (or internuclear) distance changes, the dipole moment
will change, and to a good approximation it may be assumed that the dipole
moment is a linear function of the deviation from the equilibrium position of
the interatomic distance:
D = Do + qx. (1.2)
1 This statement applies only to dipole radiation. There may be quadrupole or higher multi-
pole radiation even for molecules for which the dipole moment is zero; however, the magnitude
of the higher multi pole radiation is exceedingly small.
III.! Transitions between Energy Levels of Vibrating Molecules 119
Therefore the dipole moment changes with the frequency of the mechanical
vibration. Oscillating charges radiate an electromagnetic field, and on the
basis of classical electrodynamics the emitted light should have a frequency
equal to the frequency of the oscillator, i.e.,
OJ
V = 2n' (1.3)
where OJ = Jkjm is the angular frequency of the classical oscillator [see

(11.2.8)].
If the molecule consists of two like atoms (e.g., O 2 , N 2 ), then the dipole
moment is zero, because the centers of positive and negative charge coincide
and oscillations of the molecule about its equilibrium position do not lead to
oscillations of the center of charge. No emission or absorption of electro-
magnetic radiation occurs.
Let us now turn to the quantum-mechanical molecule. Quantum-theoreti-
cally, the emission of radiation takes place as a result of a transition of the
oscillator from a higher to a lower energy state, and absorption takes place by
the reverse process. The frequency of the emitted light is given by
(1.1)
The intensity of the emission, classically proportional to the time-averaged

value (over one period) ofthe square of the dipole moment D, is in quantum
theory proportional to the absolute value squared of the transition matrix
elements
(1.4)
where D is the dipole operator, obtained from (1.2) by the usual procedure of
replacing the number x with the operator Q:
D = Do + qQ. (1.5)
The transition probability (intensity) Anm for spontaneous dipole emission 2
in the transition from an energy state with energy value En to a state with
energy Em is given by 3
(1.6)
where OJ nm = (En - Em)/fi, c is the velocity of light, and
2 1 1~" 2
IDmnl = dim(AnYl') Tr(An D . Am D ) = dim(AnYl'L~l v':-;, I<.u, mlDdn, v)1 .
(1.7)
2 The probabilities for induced emission and absorption are also proportional to I Dmnl 2 and
to the intensity of the incident radiation.
3 This formula can be obtained using semiclassical arguments, or it can be derived by applying
the results of Chapter XXI.
D is given by (1.5), An is the projection operator on the energy eigenspace

with eigenvalue En, and dim(An £) is the dimension of this energy eigenspace.
Ii and v are the same kind of index as K in (II.4.35); they label the different
vectors within the energy eigenspaces Am£ and An£.
Since we are now not interested so much in knowing the intensity as we are
in knowing when this intensity is zero, the precise form of the transition
probability is not of primary interest to us at the moment. Equation (1.6) can
be derived using the general formalism developed in Chapter XXI. For the
special case of the one-dimensional oscillator we replace the dipole and
position vectors by the one-dimensional quantities D and Q and (1.7) goes
over into J<mJDJn)J2.
For many quantum-mechanical systems, a great majority of the matrix
elements of D vanish, so there is a severe limitation on the possibilities for
transitions. The rules that express this limitation are called selection rules.
In order to determine which particular transitions can actually occur for the
harmonic oscillator, we have to calculate the matrix elements
<mJDJn) = q<mJQJn). (1.8)
The matrix elements of the position operator between energy eigenstates have
already been calculated [see Equations (11.7.3)J and are given by
<mJQJn) = J2:W (In <mJn - I) + Fn+l <mJn + 1». (1.9)
Thus we see that the transition probability and hence the emission and
absorption intensity of light are zero except when the quantum numbers n
and m differ by unity. Thus the selection rule for the harmonic oscillator is
n-m= ±1. ( 1.10)
Transitions in the harmonic oscillator are possible only between neighboring

energy levels. The frequency of light that is emitted (for En > Em) or absorbed
(for En < Em) is given according to (1.1) and (1.10) by
En - Em flw [ 1) ( l)J W
vnm = h = h (n +2 - m +2 = 2n' (1.11)
Thus quantum-theoretically the frequency ofthe radiated light is equal to the

frequency w/2n of the oscillator, and is independent of the energy level n.
Similar arguments apply for absorption. Thus we have seen that for the
particular case of the quantum-mechanical harmonic oscillator, the frequency
of emitted and absorbed light is the same as it would be for the classical
oscillator.
If we recall the energy-level diagram for the harmonic oscillator (Figure
11.4.3) we can indicate the allowed transitions by vertical lines (see Figure 1.1).
All these transitions give rise to the same frequency. This is a consequence of
the equal spacing between energy levels.
III.I Transitions between Energy Levels of Vibrating Molecules 121
--------r.L-- E.
- - - - - -r--- EJ
_ _ _ _ _--,--'L-...._ _ El
-----,--1.--- E.
-------~~-----Eo
Figure 1.1 Dipole transitions between the energy levels of the harmonic oscillator.
For a diatomic molecule consisting of two like atoms (e.g., 02), the dipole
moment operator (1.5) is the zero operator, and therefore no transitions
between different energy levels occur.
Let us now turn to the comparison of our theoretical results with the
experimental situation.
In order to find out what frequency we should expect, we turn first to the
energy-loss spectrum of the CO molecules (Figure 11.4.2). From the distance
between the bumps in the energy-loss spectrum we find that the difference
between the various energy levels of the vibrating CO molecule is
t1E = 0.265 eY. (1.12a)
If we calculate the frequency from this according to (1.11) we find
t1E 0.265 eV
v = ~ = = 6.4 X 1013 sec-I (1.12b)
2nh 2n x 6.58 x 10- 16 eV sec
and
c
A = - = 0.466 x 10- 3 cm = 4.66 f.1 (1.12c)
v
(If.1= 10 - 4 cm; 1 A = 10- 8 cm = 10- 4 f.1).
In molecular spectroscopy it is customary to give the frequency not in
sec - 1 but in cm - I, i.e., give instead of the frequency v the wave number
vic = 1/ . 1" which indicates the number of waves per cm. We shall not intro-
duce a new symbol for it but also call it v, and the unit next to the number
will then tell us what is meant. The frequency in cm - I, or wave number, of
the radiation emitted by the transition between the vibrational levels of
CO is then
(1.12d)
Thus we expect, from the energy-loss spectrum of CO, that the vibrating CO
molecules emit or absorb electromagnetic radiation only with the frequency
given by (1.12), i.e., we expect one spectral line in the near infrared region. 4
4 In order to give a ~eeling for the orders of magnitude of various quantities involved in
molecular spectroscopy, we show a table of the regions of the electromagnetic spectrum in
Figure 1.2.
If we compare this with the absorption or emission spectrum, we find that

it is indeed correct. If the absorption spectrum is obtained with a thin layer
of absorbing gas, one finds only a single, broad, intense absorption line (or
band) in the near infrared region, with a wavelength around A. = 4.66 p.
For other diatomic molecules consisting of unlike atoms, one finds the same
situation; e.g., for Hel this band lies at A. = 2.46 p. One also finds that such
bands do not appear for molecules consisting of like atoms, such as O 2 ,
N2 , H2 •
If the absorption is observed with thicker layers of gas, the intensity of
absorption of the fundamental band naturally increases, and in addition a
second band of similar form appears quite weakly, at approximately half the
THE ELECTROMAGNETIC SPECTRUM
Energy Frequency Transitions Radiation Wave Number Wave Length
E (ev) v (sec-I) v(cm- I ) ). (cm)

Nuclear Magnetic
Resonance Radio
Spin Orientations Waves
5 x 10- 6 1.2 X 109 in Magnetic Field 4 x 10- 2 25
Electron Spin Micro-
Resonance waves
(radar)
3.1 x 10- 3 7.5 lOll Molecular 4 x 10- 2
X 25
Rotations
5 x 10- 2 1.2 X 10 13 400 2.5 x 10- 3
Molecular Infrared
Vibrations Region
0.5 1.2 x 10 14 4000 2.5 X 10- 4
1.55 3.8 x 10 14 12.5 X 103 8 x 10-'

Visible
3.1 7.5 x 10 14 25 X 10 3 4 x 10-'
Valence Electronic
Transitions
6.2 1.5 x 10 1 ' Ultra- 50 x 10 3 2 x 10-'
violet
1240 3 x 10 17 10' 10-'

Inner Shell
Electronic
Transitions X-ray's
1.24 x 104 3 X 10 18 10 8 10- 8
Nuclear Gamma
Transitions Rays
Figure 1.2 Schematic diagram of the electromagnetic spectrum. Note that the scale is
nonlinear. Boundaries between regions are generally quite arbitrary.
III.! Transitions between Energy Levels of Vibrating Molecules !23
11 = 1
11 = 2 11 =3 11 = 4 11 = 5
o 5000 10000
Figure 1.3 Coarse structure of the infrared spectrum of HCl (schematic). The intensity
actually falls off five times faster than indicated by the height of the vertical lines.
Herzberg (1966), vol. 1.
wavelength or double the frequency (wave number). If the thickness of the

layer is still further increased (up to several meters at atmospheric pressure),
a third and possibly even a fourth and a fifth band appear whose wavelengths
are approximately a third, a fourth, and a fifth, respectively, of that of the
first band; that is to say, their frequencies are three, four, and five times as
great. Figure 1.3 gives schematically the complete infrared spectrum of HCl.
In this figure the lengths of the vertical lines that represent the bands give an
indication of their intensity. However, the actual decrease in intensity is five
times as fast as is indicated in the drawing.
The explanation of these additional bands with lower intensity is that the
diatomic molecule is not quite a harmonic oscillator. In a harmonic oscillator
the restoring force increases indefinitely with increasing distance from the
equilibrium point. However, it is clear that in an actual molecule, when the
atoms are at a great distance from one another, the attractive force is zero.
Thus the quantum-mechanical harmonic oscillator is only a simplified model
of the vibrating molecule, and if one wants to describe the finer details of
vibrating molecules, then the anharmonic forces also have to be taken into
account. The energy levels of the anharmonic oscillator are not equidistant
like those of the harmonic oscillator, but rather their separation decreases
slowly with increasing n.
The energy levels and absorption spectrum for an anharmonic but almost
harmonic oscillator are shown in Figure 1.4. (For the sake of clarity a faster
decrease of I1E is drawn than is actually found in most observed cases.) The
selection rule (1.10),
n - m = ±1,
holds only approximately for the anharmonic oscillator and applies only to
the most intense transitions. But now transitions with n - m = ± 2, ±3, ...
can also appear-though with rapidly decreasing intensity. All these results
can be calculated using perturbation theory, which we shall introduce in
Chapter VIII. We describe these facts here to demonstrate that the simple
soluble quantum-mechanical models like the harmonic oscillator describe
only the principal structures of a microphysical system in nature, and cannot
be expected to describe all the details. This is not a deficiency of the harmonic
E
"6
4
t o
0------------------------
I .
Figure 1.4 Energy levels and infrared transitions of the anharmonic oscillator. The
absorption spectrum is given schematically beneath.
oscillator model but a general property of physical theories. Models are only
idealizations and cannot be expected to reproduce the experimental results up
to the last digit. The explanation of a new decimal place in an experimental
number often requires a new model, and sometimes a completely new theory.
We shall see this presently when we consider the transition frequencies in
the near infrared region in more detail, as obtained with a spectrometer of
sufficiently high resolution. The broad spectral lines for the CO molecule
around v = 2140 cm - 1 is then resolved into a number of individual narrow
lines, as shown in Figure 1.5. That is, around v = 2140 cm - lone does not
have a single line but a band, called the vibration-rotation band_ As one
sees from this figure, this band consists of a series of almost equidistant
lines, with one line missing in the center of the band_ Going out from the
gap, there are two branches, which are called the P branch (towards longer
wavelengths) and the R branch (towards smaller wavelengths). Figure 1.6
shows the same effect for the n = 1 line of Figure 1.3 for HCI.
One would expect such fine structure in the absorption or emission spec-
trum of electromagnetic radiation for the CO molecule if the energy levels
of the vibrating molecule of Figure 1.1 were split into a series of sublevels
as shown in Figure 1.7, which shows only any two neighboring energy levels
of the energy spectrum of the vibrating molecule as given in Figure L 1.
The description of such a splitting lies outside the capability of an oscillator
model. It can only mean that a state characterized by the quantum number n
is not a pure state but is in fact a mixture of states with different energies_ In
the oscillator, however, the state characterized by n was a pure state described
by a projection operator An on a one-dimensional subspace spanned by
rPn' namely An Yf_ The state of the diatomic molecule characterized by the
quantum number n must have at least as many dimensions as energy levels
(when the number of energy levels equals the dimension then to any energy
value there belongs a one-dimensional subspace or a projector on a one-
dimensional subspace)_ Consequently the oscillator model alone describes
only part of the properties of a diatomic molecule, and in order to describe
the finer details of the spectrum we have to combine the oscillator model with
Ill.1 Transitions between Energy Levels of Vibrating Molecules 125
.c:
u
C
'"
1)
0..
8 oj
N -0
.;;;
0
c
0
8
c
0
.D
....
ell
U
'-
0
-0
~ C
ell
N .D
c
.9
'o;j
'0....
I
c
.9
'o;j
....
.D
.;;:
0
0
~ ..c
-.
N
f-.
lI)
<II
::I
~
~
.c:
u 0
C N
'"
1)
N
N
III:
.. uopwosqv
IV
-
0\
--
m
::l
R branch P branch '"
~
[/l
'0
'"
~
....
~ .,.'" Pl
~ o
-,
jj [/l
'" N
o
'" ~ I~ ~ ~ ~
w
...~ '"...
N
"" 'f
N ~
Cl ~
.... o
~ ~ ~ ()
~ r~ .... :::J >=
'"
S! .... en
~
~
I~ ;S e '"
N S!
S ~
.... ~
~ ~ II '"e:
;: ...~ II~
.......~
~ '" ~
0- ~
I! s ... I~
N
;;;:
I~ !-' :0 II~ Ig
~ ;:j N ~
,:;:
?" S I~ ....
~ ~ 0-
'" N N
~ ~ i
~ ~
~~ '"
::;
..
t:; ~
...... c..
'"
8
~ ~ ~ ~ ...'"~ '"
Figure 1.6 The fundamental absorption band for HCl under high resolution. (The lines are doubled due to the presence of the two
isotopes Ce s and Cl 37 in the ratio 3: 1; we do not discuss this effect here.) [From Alpert Keiser and Szymanski (1970), with permission.]
1/ == 0
"0 B 2144 cm- ' >-<
1---;;- Gscale) n >-<
>-<
--1 1-- -< ~ 3.86 cm - '
...,
...,
"tI po
"' Iii: o
g-
.... I» ~.
:::>
~ (") O·
::r ::I
3 en
g"
-+ ?
(1)
(1)
t ;11:1 ::I
...r::r ITl
I» ::I
(1)
:::I ...,
(")
(JQ
::r '<
-, -s -. -, ., .... .... -,-......... -, -...: ..., t""'
(1)
.... .... ....
I • i i I I II I ~ " I II II II I I <i
0_", 0- '0 0 -,..,.. ....,
. c "" ~
"
3 o
.....,
Energy ground vibralion stale: [ Aboorpl ion Energy upper .,bral,on .Iale
<
0'
...,
r -
Figure 1.7 Origin and appearance of rotational structure. P and R branches are shown to the left and right, re-
spectively, on the spectrometer tracing of the CO fundamental absorption band at 2144 cm - 1. The Q branch (dashed
e
::i"
(JQ
line) is missing. Energy levels are shown to scale, except that the distance between upper and lower vibrational 3::
states (2144 cm -1) should be about five times as great as in the figure. [From Bauman (1962), with permission.] o
(p
()
~
(p
en
tv
-.J
-
another model, which describes these finer details and which reflects ad-
ditional features of the diatomic molecule that we have so far not taken into
account. This new model is the rotator model.
If we return to our classical picture of the CO molecule, considering it as
two atoms of mass m 1 and m2 at a distance x, then we see that this classical
object not only can perform vibrations along the x-axis, but also can perform
rotations in three-dimensional space around its center of mass. As long as
it is in the vibrational ground state, that is, as long as the energy involved is
less than 0.26 eV for CO, it is a rigid rotator, i.e., it can be considered as two
point like masses ml' m2 fastened to the ends of a weightless rigid rod of
length x. We shall therefore first study the rigid-rotator model by itself.
This will provide us with a description of the CO states that are characterized
by the quantum number n = 0, and will also approximately describe each
set of states with a given vibrational quantum number n. Then we shall see
how these two models are combined to form the vibrating rotator or the
rotating vibrator.
111.2 The Rigid Rotator
To conjecture the mathematical image (the algebra of operators) for the

rotator, we proceed in the same way as for the oscillator: We consider the
classical rotator, and we replace the three coordinates of momentum Pi and
the three coordinates of position Xi in all expressions for the observables with
operators Pi and Qi that fulfill the canonical commutation relations
Ii
[Pi' QJ = --:bijI,
1
[Qi' QJ = 0, [P;, PJ = (2.1) °
°
(bij = 1 for i = j and bij = for i :I j; i,j = 1,2, 3)-the obvious generaliza-
tion of (l1.2.1a) to the three-dimensional case.
In classical mechanics the energy E of rotation of a rigid body is given by
(2.2)
Here (j) is the angular velocity of the rotation, 5 and I is the moment of inertia
of the system about the axis of rotation. The angular velocity is related to the
number of rotations per second, VIOl (the rotational frequency), by
(j) = 2nv rol • (2.3)
The angular momentum of the system is given by ) = n • co. Introducing this
into (2.2), the energy may also be expressed by
)2
E = 21. (2.4)
5 We use the same symbol ill for the rotational angular velocity that we used for the vibrational
angular velocity in Chapter II.
III.2 The Rigid Rotator 129
I- x---r-------
C
Figure 2.1 Dumbbell model of a diatomic molecule.
The moment of inertia about the axis of rotation for the rotator model is
given by
where
(2.5)
are the distances of the masses ml and m 2 from the center of mass C, and x
is the distance between the two mass points (see Figure 2.1). Substitution
gives
(2.6)
that is, the moment of inertia about the axis of rotation is the same as that of
a mass point of mass
(2.7)
at a distance x from the axis; jJ. is called the reduced mass of the molecule.
Thus, instead of considering the rotation of the rigid rotator, we can equally
well consider the rotation of a single mass point of mass jJ. with coordinates
Xi' where the vector x = (Xl, X2, X3) is the position vector for the mass point
jJ. with the center of mass C as the origin. If we denote the momentum of the
mass point jJ. in this coordinate system by p = (Pt. P2, P3), then the angular
momentum about the point C is given by
I = x x p, (2.8)
so its components are given by
Ii = L EijkXjPk == EijkXjPk' (2.9)

j. k
In this equation Eijk = + 1 for (ijk) = (123) and every even permutation
thereof, Eijk = -1 for (ijk) an odd permutation of (123), and Eijk = Oother-
wise.1t is understood from now on that two identical indices in a term indicate
a summation from 1 to 3 on those indices unless stated otherwise.
l30 III Energy Spectra of Some Molecules
The operators that represent the corresponding quantum-mechanical

observables are then obtained by our general principle of replacing X;, Pj
by Q;, P j fulfilling (2.1). Thus the angular-momentum operator L is given by
L = Q x P,
or
(2.10)
and the energy operator is given in correspondence to (2.4) by

1 1 3 1
H = -V = - " LL = -L·L- (2.11)
2I 2 I ;-:-1 " 2 I ' ,.
As Qj and Pk are Hermitian operators, it follows that L; and H are Hermitian:

(2.11a)
From the Heisenberg commutation relations (2.1) one obtains the com-
mutation relations of the operators L; that represent the components of
angular momentum. The calculation goes as follows:
[L;, L 1] = f;jkflmn[QjP b QmPn]
= f;jkflmn(QlP k, QmPn] + [Qj' QmPn]P k)
= f;jkflmn(Qj{[P b QmJPn + Qm[P b Pn]}
+ {[Qj' Qm]Pn + Qm[Qj' Pn]}Pk)
= fijkflmn( QjPn~(5km - QmPk~(5jn)

h
= -;- (f;jkflknQjP n - f;jkflmjQmPk)'
I
By changing the summation indices we can write this as
We now make use of the following property of the f;mk' which is easily verified:
(2.13)
Writing the second term in (2.12) in the form (2.13):
and adding it to (2.13) gives
The last equality again follows from (2.13). Substitution of this expression
into (2.12) gives
m.2 The Rigid Rotator 131
If we insert the definition (2.10) of Li on the right-hand side, we obtain the

commutation relation of the angular-momentum operators Li :
(2.14)
We have obtained the commutation relations (2.14) of the angular-
momentum operators (2.10) from the Heisenberg commutation relations
(2.1). The expression of the energy operator (2.11) does not contain the Pi
and Qi explicitly; this is true for all physical observables of the rotator. In
fact the operators Pi and Qi are unphysical observables for the quantum-
mechanical rotator system in the following sense: It is not possible to prepare
the quantum-mechanical rotator so that it is in a generalized eigenstate of
momentum or position. Therefore, for the rotator, the Li that obey the com-
mutation relations (2.14) are the fundamental physical observables, and
the fact that they may be obtained from the momentum and position opera-
tors can be ignored. In fact, the operators Li defined by (2.10) or the classical
quantities Ii defined by (2.9) are a particular case of an observable that is
connected with new degrees of freedom of physical systems in the three-
dimensional space.
A physical object in the three-dimensional physical space has six degrees
of freedom: three translational degrees of freedom described by the three
coordinates Xi' and three rotational degrees of freedom, described by a
rotation R(!X, /3, y) that depends upon three angles !x, /3, y (e.g., the three
Euler angles, or the three angles of rotation around three fixed coordinate
axes). The momentum Pi is the canonical variable conjugate to the coordinate
Xi' and the canonical variable conjugate to the angular coordinate !Xi is the
angular momentum Ii'
If we forget about the constituents of the diatomic molecule and consider
the dumbbell as an entire physical system whose center of mass is fixed at a
particular point in space, then the remaining degrees of freedom are the
rotations by the three angles, and the dynamical variables are these angles
and their conjugates, the angular momenta Ii' In our particular case above
the Ii are obtained from the momenta and coordinates of the constituents,
which are thought of as mass points, i.e., not possessing rotational degrees of
freedom.
In general, a classical extended particle, in the three-dimensional physical
space, has as variables the momenta Pi and the angular momenta Si (spin)
associated with it, which are the conjugates of the linear coordinates Xi
and the angular coordinates !Xi respectively. For the case that the extended
physical object is a dumbbell the spin Si is the orbital angular momentum
Ii = tijkXjPk of its two constituents.
For the quantum physical extended particle, the momentum is represented
by the operator Pi' and the spin by the operator Si' It is now easy to con-
jecture that the defining commutation relations of the spin operators Si are
(2.15)
For the special case of the spin of the dumbbell, these relations were derived
in (2.14) from the relations (2.1), which had already proven to be successful.
Equation (2.15) can also be derived from the properties of the group of rota-
tions, if one assumes that the rotation R(a, (3, y) of the physical object is
represented (continuously) by a (unitary) operator U(a, (3, y) in the space
of physical states of this quantum-physical object. This latter assumption
can be derived from the fact that the rotations are symmetry transformations
and that the rotation group is a symmetry group of the physical system
(Wigner Theorem). We will not discuss this connection here, but we take
the algebra of observables defined by (2.15) as the starting point.
We will now investigate the properties of the algebra of operators generated
by the J i , which fulfill the commutation relations
(i, k, I = 1, 2, 3), (2.16)
where J i may be either an Li as given by (2.10) or an Si' In particular we shall
find the properties of all the J i , which are linear Hermitian operators in a
linear space. It will turn out that the set of all J i is richer than the set of the Li
given by (2.10). The algebra generated by the J i is called the enveloping
algebra of the group SU(2) and is denoted by 6"(SU(2).6
I1L3 The Algebra of Angular Momentum

We shall now find all possible solutions of the commutation relation (2.16)
that fulfill Jl = J i' This means that we shall construct all linear spaces in
which the J i that fulfill (2.16) act as linear Hermitian operators. We assume
that there exists at least one eigenvector of J 3 in these spaces.
Instead of working with the J i (i = 1,2, 3), we introduce the following
linear combinations:
H3 = 11- 1J 3, H+ = 11- 1 (J 1 + il 2 ), H_ = 11- 1 (11 - il 2 ), (3.1)
The Hermiticity condition Jl = J i is expressed by
(3.2)
From (2.16) it then follows that
[H3,H±J = ±H b
[H +, H _ J = 2H 3' (3.3)
The operator J2 can be written
(3.4)
with
6 J; are called the generators of the group SU(2), meaning that every element of the group
is given by e;"JJJ. The commutation relations of the group SO(3) and the group SU(2) are the
same. We therefore use the symbol 8(SO(3)) and 8(SU(2)) interchangeably. In general, the
enveloping algebra of any Lie group (of operators) is defined as the associative algebra whose
generators---in the sense of the definition of Equation (I.3.13)-are the generators of the
Lie group.
H2 and therfore J2 have the property that

(3.6a)
or in general
(3.6b)
where A is any element of the form
A = aI + aiJ i + a ijJ;Jj + a ijk J;JjJ k + "', (3.7)

where a, ai, aij, aijk, ... , are complex numbers. Because of the property (3.6),
H2 and J2 are called invariant operators (Casimir operators) of the algebra
t5"(SU(2)).
We now find all ladder representations of t5"(SU(2)), i.e., all solutions of the
commutation relations (2.16) by linear operators in linear spaces that are
obtained by applying to one eigenvector of H 3 the whole algebra t5"(SU(2)).
(This one eigenvector is assumed to exist.)
We choose a vector f = r which is an eigenvector ofH 2 with eigenvalue c:
(3.8)
As H2 commutes with every A of the form (3.7), Afis again an eigenvector
of H2 with eigenvalue c.
o
We now choosefsuch that it is also an eigenvector of H3 and we call the
eigenvalue m. The normalized eigenvector we denote by fm = f':n :
(3.8')
(j~are called weight vectors; m is called a weight.)

To choose a vector such that it is simultaneously an eigenvector of two
operators is possible iff these two operators commute, because then
H2H3fm = H3 H2fm
which by (3.8) and (3.8') becomes an identity. Note that if two operators do
not commute then, in general, they have no com~n eigenvector (see
Problem 7).
Now we define
fm± = H ± fm
and calculate using (3.3)
H3fm± = H3 H ±fm = (H±H3 ± H±)fm

= (H±m ± H±)fm = (m ± I)H±fm
= (m ± l)fm±'
(3.9)
Thus Im± is, if it is different from 0, an eigenvector of H 3 with eigenvalue

mil.
We note some properties offm:
1. The number c is nonnegative.
PROOF.
c = Um, H2fm) = h- 2 Um, (Ji + J~ + JDfm)
= h- 2(IIJdmI1 2 + IIJzfml1 2 + IIJ 3 fm112) 2 o. o
2. Any eigenvalue m of H 3 fulfills
m2 ::s; c. (3.10)
PROOF.
0:0; h- 2{(Jdm, Jdm) + (Jzfm, Jzfm)}
= h- 2{Um, J 2fm) - Um, J~fm)} = h- 2h 2(c - m2)llfmI1 2,
from which follows (c - m2 ) 2 O. 0
Note that in the proofs the Hermiticity property of J i has been used.
If we start with an arbitrary eigenvector I'mo of H 3 and H2 and apply
H + successively, then we obtain according to (3.9) new eigenvectors I'm of
H 3 with ever-increasing eigenvalues. Because of (3.10), after a finite number
of steps we must reach the eigenvector f! with the largest eigenvalue I of
H 3, i.e., with
f!+ = H + f! = O. (3.11)
H2f! = (H~ + H 3 )f! = l(l + 1)f!. (3.12)
Consequently the eigenvalue c of H2 and the largest eigenvalue I of H 3
(highest weight) are connected by
c = l(l + 1).
Instead of characterizing the eigenvectors of H2 and H 3 by c and m, we can
characterize them by I and m.
If we apply H _ successively to I~, then after a finite number of steps,
because of (3.10), we must reach the vector I~ with the lowest eigenvalue f1
of H 3, i.e., with
I~- = H-I~ = O. (3.13)
H2I~ = (H~ - H3)I~ = f1(f1 - I)I~· (3.14)
Comparing (3.14) with (3.12) we find
1(1 + 1) = f1(f1 - 1),
and the only solution of this equation for f1 that fulfills (3.10) is
f1 = -I. (3.15)
Thus if we start with the vector fl, apply H _ successively to it and normal-
ize, we obtain the sequence of vectors
fl-l = (cx/)-IH_fL
fl-z = (CX/-1)-IH_fl-l,
(3.16)
where the CXm are

CXm = J(H - f~, H - f~)·
We shall finally reach the vector with the smallest weight /1:
f~ = (cx lL +I)-I H - f~+ I·
As /1 = -I, we have 21 + 1 vectors in the sequence (3.16).
f~ with m = I, 1 - 1, 1 - 2, ... , - 1 + 1, -I, (3.17)
which fulfill
(3.18)
As 21 + 1 is a number of vectors, it must be an integer; consequently 1can be
only one of the following numbers:
1 = 0, t, 1, t .... (3.19)
Thus for a given number I, which can be one of the numbers in (3.19), we
have 21 + 1 vectors f~ that are orthogonal to each other and span a space,
which we shall call ~l :
(3.20)
We shall now determine the value of the normalization constant cxm:

CXmCX m = (H-fm,H-fm) = Um,H+H-fm)
= Um, (HZ - H~ + H 3 )fm) = l(l + 1) - mZ + m.
Thus, except for a phase factor (which remains undetermined),
CXm = JI(l + 1) - mZ + m = J(l + m)(1 - m + 1), (3.21)
and we have
H - f~ = J(l + m)(1 - m + 1) f~-I = CXmf~-I. (3.22)

It remains to determine H + fm. We know already that H + f~ ex f~+ I·
We set H + f~ = f3mf~+ I and calculate
f3mU~+ I, f~+ I) = (H + f~,f~+ 1)"= U~, H - f~+ I) = CXm+IU~,f~).

-..----••
-2 -1
----••0----••
+1
----~.~--~~m
+2
1=2
Figure 3.1 Example of a weight diagram of an irreducible representation of SU(2).
Therefore
Pm = ~m+l = J(l + m + 1)(/- m) = Pm'
and we have
(3.23)
Summarizing, we have found the following: For every integer or half-

integer value 1 there is a space ~l spanned by 21 + I orthogonal vectors
f~ (m = -I, ... , I). In this space the operators H +, H _, H 3 are given by
(3.8'), (3.22), (3.23); and thus the action of any element A E !&"(SU(2) given
by (3.7) on any vector f E ~l given by (3.20) is determined. To indicate that
for every lone obtains a different operator, one could also write H<J!, HI!!,
H~) for the operators in (3.8'), (3.22), (3.23). The space ~l is called an irreduc-
ible representation space of !&"(SU(2». In this space the elements H + , H _ , H 3
defined by (3.3) are represented by the operators given in (3.8'), (3.22), (3.23).
These operators are called the (21 + I)-dimensional irreducible representation
of the H +, H _, H 3' As can be seen, they depend upon I; for each I = 0,
to 1, ... , there is one" different" set of operators. All of the vectors in ~l are
eigenvectors of H2 with the same eigenvalue l(l + 1). There is no element
A E !&"(SU(2» that can transform from a vector l E ~l to a vector fl' E ~l'
with t -=I- 1'. This fact is expressed by the statement that ~l is "invariant
under all A E !&"(SU(2»." In particular, ~l remains invariant under the action
of the J i (i = 1,2,3) and the H ±, H 3 .
If we plot the possible values of m in an irreducible representation along a
line, then we obtain the weight diagram of the representation characterised
by t, ofSU(2). This is the simplest example of a weight diagram for Lie groups.
For I = 2 this is shown in Figure 3.1. To each dot there corresponds a basis
vector f~ in the representation space ~l, or equivalently there corresponds
the one-dimensional subspace spanned by f~. Each such subspace (or basis
vector) represents a (pure) physical state. Thus to each point in the weight
diagram there corresponds a pure physical state.
The smallest space is ~o, which is one-dimensional, as shown in Figure
3.2.
In general, we have a weight diagram for each representation, and the
action of the operators H + , H _ can be represented in this diagram, as shown
--------------~.~------------.~m
o
1=0
Figure 3.2 Weight diagram of the one-dimensional representation of SU(2).
• • • • • • gp5i2
• • 2 • • al 2
H_
al'12
• • •• •
H+ 9i'1
•
gpl 2
• •
0 m dfo
-3 - 2 - 1 0 +1 +2 +3
Figure 3.3 Weight diagrams of irreducible representations of SU(2).
in Figure 3.3. For each weight diagram, there is a space ~I, and for each space
~I there is a possible state (or set of states) of the quantum-mechanical
system characterized by the value I. Because of the correspondence between
the operator L (2.10) and the classical angular momentum I (2.8), the number
I is called the angular momentum quantum number:
Thus the quantum-mechanical angular momentum can take only a discrete

number of values. The physical state that corresponds to a given space ~I,
and which is described by the statistical operator W = (dim ~')-IAI =
°
(21 + 1)-1 N where N is the projector onto ~I, has a definite angular mo-
mentum I. Except for I = this is not a pure state but a mixture. The prepara-
tion of a pure state requires not only a measurement of L 2 but also a measure-
ment of L3 (or any other component of the vector L). If the value of L3 is
always measured to be m then the state is the pure state represented by the
projection operator A~ = If~> <f~1 on the one-dimensional subspace ~~.
The space ~I is the direct sum of one-dimensional spaces ~~,
+1
~I = I EPJ ~~, (3.24)
m~ -I
and each ~~ is spanned by the vector f~.

It has already been mentioned that not all sets of linear operators J i
(i = 1, 2, 3) that fulfill (2.16) can be given by (2.10). It can be proven (Problem
1) that for Li given by (2.10) the number I can only take the values I =
0,1,2, ... ; in other words, the operators Li given by (2.10) can be represented
by operators in the spaces ~I with I = 0, 1, 2, ... only. Thus for the operators
given by (2.10) there are a countable number of representatives Lj') in spaces
~I that remain invariant under the action of the Ljl). These spaces ~I are not,
however, invariant under the action of the operators Qj and Pj. Using (2.1)
and the definition (2.10), one calculates immediately
[L i , Qj] = ih£ijkQb
and
[L i , Pj] = ih£ijkPk. (3.25)
Operators like P and Q, which fulfill these commutation relations with L,
are called vector operators. From (3.25) it follows that
[L2, Qj] = ih£ijk(LiQk + QkLi) =I 0,
so that Qj will change the eigenvalue 1(1 + 1) of L2. This means that Qj can
transform from one PAL to another PAL' with l' =I l.
The operators Li in PAL with 1 = t, !, ~, ... or any direct sum of them,
L EB PAL
I = half-odd
integer
cannot be expressed as functions of the operators Qi and Pi. For 1 = the t

operators L!I= 1/2) are called spin operators. The 2 x 2 matrices (Ii with the
matrix elements 2(f~=' 1/2, LJ~= 1/2) are called Pauli matrices (Problem 5).7
111.4 Rotation Spectra

As mentioned before, the algebra t9'(SU(2)) does not contain operators that
transform out of a given PAL. The algebra of observables of the quantum-
mechanical rotator is, however, larger than t9'(SU(2)); additional elements can
be formed, e.g., algebraic functions of J i and Pi or of J i and Qi. The observ-
abIes Qi, for instance, have the property that they transform from a given
PAL into the neighboring PA l + 1 and PAL-I:
Qi: PAL --+ PA I - 1 EB PAL + 1, (4.1)
but not into PAI±n for n > 1. We shall prove this in the section on parity.
The space of physical states f7t of the quantum-mechanical rotator is the
direct sum of the spaces PAL :
co
PA = L EB PAL. (4.2)
1=0
7 The representations of the J, given by (3.8'), (3.22), (3.23), and (3.1) for integer values of I
are related to group representations of the rotation group SO(3) by taking e'o,') I, e'Q)21" e'Q)'),
and products thereof, where e'Q)'), represents the rotation around the third axis by an angle w3 .
For half-odd integer values of I the J, are not related to group representations ofSO(3); they are,
however, still ray representations of SO(3). As physical states are represented by rays and not
by vectors-according to hypothesis II-not only proper group representations but also ray
representations are relevant for quantum physics. Therefore representations with any integer
or half-odd integer value of 1are realized in quantum physics. All these representations are group
representations of the "covering group" SU(2) which, therefore, has also been called (by E. P.
Wigner) the quantum mechanical rotation group. The occurrence of half-odd integer angular
momenta is, therefore, a natural consequence of the basic assumptions of quantum mechanics.
In terms of the algebra of observables this means that all representations of the algebra of angular
momentum (2.16) are realized in nature.
I1I.4 Rotation Spectra 139
Before we justify this statement, we want to give a brief description of the

properties of [lll.
[lll is not an irreducible representation space of S(SO(3)), the algebra of
angular momenta. It is called a reducible representation space, and it reduces
as given in (4.2) into a direct sum of irreducible representation spaces [llli
(l = 0, 1, 2, ... ). The operators H + , H _, H 3, now considered as operators
in the big space [lll, transform every element of a given space [llli into an
element that is again in the same [llli. Thus the subspaces [lJll of [lll are left
"invariant" by the H+, H_, H 3 , and consequently by any A ES(SU(2)).
The operator H2-which is a number in the space [lJll, namely l(l + 1)-has
in [lll a notrivial spectrum, namely
spectrum H2 = {I(l + 1), I = 0, 1,2,3, ... } (4.3)
The weight diagram for the representation in [lJl is shown in Figure 4.l.
We shall now justify the statement (4.2). We assume that the J i are the
angular momenta Li = fijkQjPk; then, according to the result stated in
Section 111.3 and proved in Problem 1, only integer values of I are allowed,
i.e., [lJl contains only [lJll with I = 0, 1, 2, .... That [lll contains all the .ljfl
(l = 0, 1, 2, ... ) follows from the fact that there are observables for the rotator
(e.g., the operators Qi) that transform from a given [llli to the neighboring [llll- 1
and [llll+ 1 according to (4.1). That each [llli appears only once follows from
the fact that for the rotator no additional quantum number is necessary; if
one [lJllo were to appear twice or more, then there would be two or more
vectors f~(1), f~(2), ... with the same quantum numbers 1o, m, and a new
quantum number would be necessary to distinguish between these two or
more vectors. But the rotator is just that model for which there is no other
diagonal observable besides angular momentum (L 2 and L3)' (In other words,
a real physical system can be a rotator only to the extent that no other quan-
tum numbers are necessary for the description of its properties; e.g., poly-
atomic (symmetric top) molecules cannot in general be described by the
rotator model, and even for the diatomic molecule, the rotator model is only
an approximate description that neglects all but the rotational properties of
• • • 3 • • •
• • 2 • •
• •
0 rn
- 3 - 2 - I 0 +1 +2 +3
Figure 4.1 Collection of SO(3) weight diagrams belonging to an irreducible representa-
tion of SO(3,l) or E(3).
the dumbbell.) Thus, as always, the justification of (4.2) is that in nature there
are physical systems whose physical states are (up to a certain limitation)
described by 9£.
[Mathematically this can be stated more briefly: "The spectrum-
generating algebra of the rotator is 0"(E3)'" 0"(E3) is generated by
Lb Qi that fulfill the commutation relations
[L;, L j ] = ilifijkLb [Li' Qj] = ilifijkQk' [Qi' QJ = 0,
and :Jlt is a particular irreducible representation space of 0"(E3)']
Each dot on the weight diagram of 9£ represents the pure state that is
described by the one-dimensional subspace :Jlt~ (I, m fixed) spanned by f~.
The (normalized) statistical operator for the pure state W = A~, where A~
is the projector on :Jlt~, represents a quantum-mechanical system for which
the angular momentum has the definite value 1 and the 3-component of
angular momentum, H 3, has a definite value m. As there is no distinguished
direction in space and the coordinate system has been chosen arbitrarily,
H 3 represents the angular momentum around an arbitrarily choosable
direction; it is also called the helicity.
The values of the energy operator in 9£, i.e., the energy spectrum of the
rotator is obtained from (2.11) as
1
spectrum H = E/ = 21 1i 2 1(1 + 1). (4.4)
Thus we see that the energy levels depend upon I, as represented in the
diagram of Figure 4.2. If we compare this with Figure 1.7, we see that the
E J
1
1 6
(a)
L..:::c::::::======= ~
(b) I II I II I I I I
o
Figure 4.2 Energy levels and infrared transitions of a rigid rotator: (a) The energy-
level diagram, (b) the resulting spectrum (schematic). [From Herzberg (1966), with
permission.]
rotator has indeed the energy spectrum required to interpret the infrared
spectrum of diatomic molecules.
In contrast to the oscillator, the energy eigenspaces (i.e., the spaces of
vectors with the same energy eigenvalue) for the rotator are not one-dimen-
sional, except for I = O. Therefore the state of a rotator with a definite
energy value E,o (10 =f. 0) need not be a pure state. If only energy measure-
ments have been made with the result Elo ' the statistical operator is given by
W = N° (unnormalized), (4.5)
or
W = (Tr NO)-1 Ala = ( 2/ 0 + 1)-1 N° (normalized), (4.5')
where N° is the projector on the (2/0 + 1)-dimensional space ~Io. By an
energy measurement alone it is not possible to prepare a pure state of the
rotator. Only under certain additional conditions-if a direction in space is
distinguished (e.g., by an external magnetic field)-can one prepare a state
with definite helicity, i.e., a pure state A~o' If only the energy of the rotator
has been measured but not the helicity, then the states with different helicities
are assumed to appear with equal weight, which is why one chooses
W = N° = N5'/o + A~/o+1 + ... + A~ + ... + Aloo (4.6)
for the (unnormalized) statistical operator. Equation (4.6) is a special case of
(11.4.32) and (11.4.49).
In order to calculate the frequencies that can be emitted and absorbed by
a rotator, we have to know the selection rules. In our classical picture of the
rotator we can consider it as a rotating dipole moment D, with
D = const Q, (4.7)
where Q is the vector between the centers of positive and negative charge.
Classically the radiation is then a consequence of the rotation ofthis electrical
dipole moment. Quantum-mechanically, the intensity of the absorbed or
emitted radiation is proportional to the absolute value squared of the matrix
element of the operator D, i.e., proportional to
(4.8)
Thus dipole radiation will only be obtained in transitions from states
f~' to f~ between which the matrix element (4.8) is nonzero (quadrupole
and higher-order radiation is negligibly small). It will be shown in the
section on parity that [as was stated in (4.1)J
<f~,IQdf~) = 0 unless 1= l' ± 1. (4.9)
That is, the selection rule for dipole radiation of the rotator is
III = I - l' = ± 1. (4.10)
If we compare this result with the experimental situation for CO, depicted
in Figure 1.7, we observe complete agreement. Figure 1.7 shows the trans-
itions between states not only with different values of angular momentum I
but also with different values of the vibrational quantum number n.
We also expect radiation from transitions between different rotator states

of the diatomic molecule that belong to the same oscillator state n = O.
These transitions (absorption) are indicated by the i in Figure 4.2. The
frequency of this radiation, given in wave number units cm - 1 obtained by
dividing (Ll) by c, is
El' - E,
v,', = 2nl1c .
With (4.4) and (4.10) we calculate
112 (I + 1)(1 + 2) - (I + 1)1
v - - ---------
1+ 1.1 - 21 2nl1c
11
= ~ 2(1 + 1) = B2(l + 1), (4.11)
4ncI
where
(4.11')
Thus the spectrum of a simple rigid rotator consists of a series of equidistant

lines, as schematically drawn at the bottom of Figure 4.2.
We expect the frequency for pure rotational transitions to be much
smaller than the vibrational frequency because the spacing between the
rotational energy levels is much smaller than between the vibrational energy
levels as seen in Figure 1.7 (note the scale factor of! there).
The pure rotation spectrum lies in the far infrared. The absorption spec-
trum of HCl in the far infrared has been measured, and the experimental
results are given in the second column of Table 4.1. From (4.11) we expect
that the frequencies will have an equidistant spacing. Therefore in the third
column of the table the differences between the successive frequencies are
given. According to (4.11) this difference must be
,1.v = V'+I,1 - V',,-l = 2B. (4.12)
The first eleven frequencies have roughly equal spacing and fitting them with
(4.11) we obtain
BHCl = 2h ::::::1O.35cm- 1. (4.13)
8n cI HC1
The fourth column of the table gives the values calculated from (4.11) with
the value (4.13). We observe fairly good agreement between the calculated
and observed values if we compare only the first eleven frequencies in column
two with column four. The agreement becomes worse however at the higher
values of I. The differences between successive frequencies at the higher values
of 1are smaller than at the lower values of I. The frequencies are not equally
spaced but the trend suggests that (4.11) can be modified in order to obtain
better agreement at the higher values of 1as well. The last column of the table
gives a fit to the expansion
Vl+ 1,1 = 2b(1 + 1) - 4d(1 + 1? (4.14)
Table 4.1 Absorption spectrum of HCl in the far infrared. [Data for I = 1,2,3 from
McCubbin, J. Chern. Phys. 20, 668 (1952); for 1= 4, ... , 11 from R. L. Hansler and
R. A. Detjen, J. Chern. Phys. 21,1340 (1953); for I = 17, ... ,33 from J. Strong, Phys.
Rev. 45, 877 (1934).] The units of v are cm -1.
Veale = Veale = 20.881

V'-i,lobs ~vobs 20.701 -0.001837/ 3
1 20.8 20.70 20.87

20.8
2 41.6 41.41 41.74
20.9
3 62.5 62.11 62.58
20.53
4 83.03 82.82 83.39
21.10
5 104.13 103.52 104.15
20.60
6 124.73 124.22 124.86
20.64
7 145.37 144.93 145.50
20.52
8 165.89 165.63 166.07
20.34
9 186.23 186.33 186.54
20.37
10 206.60 207.04 206.92
22.26
11 228.86 227.74 227.19
17 345.6 351.96 345.86
18 364.6 19.0 372.67 365.05
19 384.2 19.6 393.37 384.04
20 403.4 19.2 414.07 402.82
21 422.1 18.7 434.78 421.37
22 440.1 18.0 455.48 439.70
23 458.3 18.2 476.18 457.79
24 474.4 18.1 496.89 475.62
25 494.3 19.9 517.59 493.18
26 511.2 16.9 538.30 510.48
27 527.7 16.5 559.00 527.48
28 542.9 15.2 579.70 544.19
29 561.1 18.2 600.41 560.59
30 576.4 15.3 621.11 576.66
31 589.6 13.2 641.81 592.41
32 609.4 19.8 662.52 607.82
33 623.4 14.0 683.22 622.87
(b, d are constants). Comparing this last column with the observed values in
the second column, we see that the agreement of (4.14) with the experimental
data is far better than that of (4.11). The energy spectrum that corresponds to
(4.14) is given byB
(4.15)
(b, d are constants). The energy levels (4.15) have been drawn in Figure 4.3
with an exaggerated value of d.
8 The energy spectrum

EI = 2n1ic[b' - d'l]/(l + 1),
which agrees well with nuclear rotation spectra, gives poorer agreement with molecular rotation
spectra than (4.15).
J
- - - - - - - 14
- - - - - - - lJ
--------------- 12
- - - - - - - - - - - - - - - - II
- - - - - - - - - - - - - - - 10
------ --------- 9
------r .-.--.. 8
••••• 7
6
4
32
10
Figure 4.3 Energy levels of the nonrigid rotator. For comparison, the energy levels of
the corresponding rigid rotator are indicated by broken lines (for J < 6, they cannot
be drawn separately). [From Herzberg (1966), with permission.]
The explanation for the better fit of (4.15) to the experimental values is
that the diatomic molecule Hel is not exactly a rigid rotator. The bonds
between atoms are not rigid, and the interatomic distance varies with the
speed of rotation, giving rise to a centrifugal distortion. Equation (4.15) can
be obtained if we return to the classical picture in which the molecule is
considered as two hard spheres (atoms) joined, not as in Figure 2.1 by a
rigid rod, but by a spring. If the molecule rotates about an axis perpendicular
to this spring, then at equilibrium the centrifugal force F /(I1X3) equals the
centripetal force k(x - xe), where k is the spring constant and Xe the inter-
atomic distance of the stationary molecule. Thus
F
k(x - xe) = -3. (4.16)
I1 X
The energy of this system is [cf. (11.2.3) and (2.4)]
(4.17)
Making use of the expansion
X2 = Xe2( 1 + X - Xe
2~+ ... ) (4.18)
and (4.16), one obtains for E

1 1
E= - F - - - (F)2 + O((F)3). (4.19)
211X; 2112kx~
The first term is the energy of the rigid rotator, and the second term is the
contribution due to the centrifugal forces. Going to the quantum system
operator by replacing the number F with the operator L2, one obtains the
energy operator
1 2 1 2 2
(4.20)
H = 2---Z L - - 2k
26 (L ) ,
J.1.X e J.1. Xe
from which the spectrum (4.14) follows. The better fit of (4.14) to the experi-
mental values confirms the above classical consideration. But we also observe
that the empirical value of d is orders of magnitude smaller than that of b,
obtained from the above fit:
bHC! = 10.438 cm -1, d HCl = 0.0046 cm - 1. (4.21)
This shows that the rigid rotator is a remarkably good model of the rotating
diatomic molecule. 9
As we shall see below, the spacings between the levels of the rotating CO
molecule are considerably smaller than for the HCI molecule. Therefore the
pure rotation spectrum of CO lies at a considerably longer wavelength,
where experimental investigation is very difficult.
We now want to obtain some quantitative features of the classical picture
for the diatomic molecule. From the value (4.13) we calculate the moment of
inertia of HCI:
With
mC! = 35.45 g = 6.0 X 10- 23 g

NA
and
1.008
mH = - - = 0.167 X 10- 23 g,
NA
one calculates (2.7):
From (2.6) we may calculate the internuclear distance of the HCI molecules
using the values I HCl and J.1.HCl :
XHCl = 1.29 X 10- 8 cm
9 The data of Table 4.1 are accurate enough to test the correction term d(l(/ + 1))2 We have
tried to find other empirical formulas like E/ = 2nlic(h' - d'l)/(/ + 1) and found that it gives a
much poorer agreement with the data of Table 4.1 than (4.15).
Thus we have calculated from the infrared absorption spectrum that the
size of the molecule is ofthe order of 10 - 8 cm. This order of magnitude agrees
very well with the values of atomic and molecular radii obtained from other
classical considerations. We want to stress, however, that x is the value for
the classical picture of the quantum-mechanical system and is not the expec-
tation value of a quantum-mechanical observable.
111.5 Combination of Quantum Physical Systems-

The Vibrating Rotator
We shall now combine the quantum-mechanical rotator model with the

quantum-mechanical oscillator model to form the quantum-mechanical
vibrating rotator (or rotating oscillator) model. This will provide a descrip-
tion of the experimental situation shown in Figures 1.5, 1.6, and 1.7. We first
discuss the general case of a combination of two quantum physical systems.
For this we require a new mathematical notion: the direct product or tensor
product of linear spaces.
[Let R I and R 2 be two linear spaces, let Ui E R I and vj E R 2, and let

aij E C (complex numbers). The set of all (arbitrarily large but finite)
sums
f = Ii,j aijUiVj, (5.1)
where aij takes any value in C, forms a linear space, which is called
the direct-product space and is denoted RI ® R 2 . UiVj is the formal
product, which is also written uivj = ui ® Vj' If (U, U')I denotes the
scalar product in R I , and if (v, v'h denotes the scalar product in R 2 ,
then the scalar product in RI ® R2 is defined by
(Iij aijUiVj, I1m blmUiV~) = I

ijlm
aijblm(uj, uiMVj, v~h· (5.2)
(Remark: If RI and R2 are Hilbert spaces, then the "completion"

of RI ® R2 with respect to this scalar product is the direct Hilbert-
space product.)
If <Pv is a basis in RI and", /l is a basis in R 2 , then
(5.3)
is a basis in RI ® R 2 •
If Al is a linear operator in RI and A2 is a linear operator in R 2 ,
then the operators
C = A1 ® I, (5.4)
III.S Combination of Quantum Physical Systems-The Vibrating Rotator 147
in R l @ R2 are defined in the following way:

Cf = L:aij(A1Ui) @ Vj' (5.4')
ij
Bf = L aiju i ® (A2 v).

ij
The linear operator A = Al @ A2 in R l @ R2 is defined by

A(uj@ Vk) = (AI @ A 2)(uj ® Vk) = (Alu) @ (A2 Vk)' (5.5)
dcf
It is easily seen that if AI, Bl are linear operators in Rl and A 2 , B2

are linear operators in R 2 , then
A1B l @ A2B2 = (AI @ A 2)(B l ® B2)' (5.6)
Every operator A in the direct-product space is a linear combina-
tion of direct products of operators, i.e.,
A = LAY) @ A~), (5.7)
i
with AY) linear operators in Rl and A~) linear operators in R2']

With the notion of the direct product of spaces we can formulate the basic
assumption about the physical combination of two quantum-mechanical
systems:
IVa. Let one physical system be described by an algebra of operators,

.Sill , in the space R l , and the other physical system by an algebra .Sil2 in R 2 •
The direct-product space Rl ® R2 is then the space of physical states of
the physical combination of these two systems, and its observables are opera-
tors in the direct-product space [given in the form (5.7)]. The particular
observables ofthe first system alone are given by Al ® I, and the observables
of the second system alone by I @ A2 (I = identity operator).
We reemphasize that IVa is a basic assumption of quantum mechanics and

can only be justified by the fact that such physical systems exist.
We shall now apply this basic assumption IVa to the diatomic molecule
that vibrates and rotates.
We called the space of physical states of the oscillator Yf. In Yf we intro-
duced a basis of eigenvectors of the operator N or HoSt. :
basis of Yf: <Pn = In) (n = 0, 1, 2, ... ). (5.8)
The action of all observables (all elements of the algebra of observables of
the quantum-mechanical oscillator) of the harmonic oscillator on the basis
vectors I n) is known from Chapter II.
We called the space of physical states of the rotator [l). In f7l we introduced
a basis of eigenvectors of the operators L3 and L2 or H rot :
basis of f7l: f~ = 11m)
(l = 0,1,2, ... ; m = integer with -I::::;; m::::;; +/). (5.9)
The space of physical states of the vibrating rotator is, according to IVa,
the direct-product space
(5.10)
and the observables are the operators Ii A~i~c ® A~~I' where A~i~c is any
observable of the oscillator and A~~t is any observable of the rotator. The
basis system in 6 is obtained as the direct product of the basis systems ¢n
in:lt andf~ in flit, and is denoted by Inlm):
Inlm) = In)® 11m) = ¢n®f~" (5.11)
We have already mentioned that the rotating diatomic molecule is not a
rigid rotator and the vibrating diatomic molecule is not a harmonic oscillator.
Furthermore, the rotations and vibrations are not independent motions of
the molecule: In our classical picture the diatomic molecule is a system of
two mass points which are connected by a massless spring. Consequently
there are interactions between the vibration and rotation caused, e.g., by
the fact that during the vibration the internuclear distance x = (X i XY/2
changes, and consequently the moment of inertia I = fJ.X 2 changes.
For the moment we want to neglect all these finer details and consider
the idealized system that is simultaneously a rigid rotator and a harmonic
oscillator-keeping in mind, however, that this is an idealized system, which
can at best be only approximately correct.
The energy operator of this idealized physical combination of the harmonic
oscillator and rigid rotator is given by
(5.12)
where
1 1 2 fJ.W 2 2
+ 21) + -2- Q
Co
Hose = rlw(N = 2fJ. P
and
H rot =~v
21 .
If the interactions between the two systems are neglected, all observables are
given by
(5.13)
From (5.12) we obtain the energy spectrum of the idealized vibrating
rotator:
1
spectrum H = Enl = hw(n + !) + 21 h2 1(1 + 1). (5.14)
The experiments show that the system constants w for the oscillator and I for
the rotator fulfill hw > h 2 /(21) (the pure vibrational transitions are in the
near infrared and the pure rotation transitions are in the far infrared).
1 0 --
I II
10-- s= 4
I
10':=::: S~
I
10==S-
I
10--S-
------------0
Figure 5.1 Energy levels of the vibrating rotator. For each of the first five vibrational
levels, a number of rotational levels are drawn (short horizontal lines.) [From Herzberg
(1966), with permission.]
The energy-level diagram that we obtain under these conditions from (5.14) is
shown in Figure 5.1. To obtain the transition frequencies we use the selection
rules
i1n = ±1 and i11 = ±1 (5.15)
given by (1.1 0) and (4.10). If we consider a particular vibrational transition
from n to n + 1 (absorption) or n + 1 to n (emission), we obtain from (5.14)
the frequencies (in cm -I)
= Vo + 2B + 2BI (i11 = + 1), (5.16R)
En+l /-1 - En I
Vp = ' ,
2nhc
= Vo - 2Bl (i11 = -1) (5.16P)
for absorption, where B is given by (4.11');
h h
B=--=- (4.11')
8n 2 cI 4ncI
and
w
Vo = 2nc'
Thus we have two series of equidistant lines, which are called the Rand P
branches, with a gap at Vo (as i11 = 0 is excluded by the selection rule). The
corresponding transitions are indicated in the energy-level diagram, Figure
5.2. The frequency spectrum calculated from (5.16) is depicted in strip (b).
I'
10
7
6
I I i I II
I
I
I
I
I
/"
,, 10
I
I 9
I
I
,
I
,
, 6
, S
I
4
3
21
/ .,--_ _ _R_ _---. p,
o
9 8 7 6S 4 3 2 1 0 1 2 3 S4 6 7 8 9 10
~~I~"~It-I-::.1-::.1t-'tl ;1=4':::t1~1+.'t:'t1tJ:;2.~1:;.1~J= (a)

~-r~f~'1~-'(....L>(....L(....Lf....Li ....Li....Li....!....LI-Li-LI-LiJ..(J..fJ..(J..(J..rJ..'!,_'_ (b)
109 8 7 6 S 4 3 2 1 '0 - 1-2- 3-4 - S-6-7 - 8 - 9-IO
Figure 5.2 Energy-level diagram explaining the fine structure of a rotation-vibration

band. In general, the separation of the two vibrational levels is considerably larger
compared to the spacing of the rotational levels than shown in the figure (indicated by
the broken parts of the vertical lines representing the transitions). The schematic
spectrograms (a) and (b) give the resulting spectrum with and without allowance for
the interaction between rotation and vibration. In these spectrograms, unlike most of
the others, short wavelengths are at the left. [From Herzberg (1966), with permission.]
The observed spectrum from Figure 1.7 is depicted in strip (a). Thus the
prediction of the vibrating-rotator model without interaction between
vibration and rotation agrees rather well, but is not quite accurate. The
observed lines in the R branch draw closer together, and those in the P
branch draw farther apart, than the predicted equidistant lines. This is due
to the interaction of rotation and vibration.
If one assumes that the moment of inertia is different in different vibrational
states, then one obtains
1
Enl = hw(n + !) + 2I h 2 1(l + 1) (5.17)
n
instead of (S.14), and for the wave numbers of the resulting lines
v= En'l~~h:n"l" = vo(n' - nil) + Bn,I'(1' + 1) - Bn"/"(/" + 1), (S.18)
where
h
Bn = 8n 2 cI n .
From this one obtains the absorption frequencies (in wave number units
cm - 1) for transitions nil I" --> nil' between neighboring vibrational levels :
vR = vo(n' - nil) + 2B n, + (3Bn' - Bn,,)1 + (Bn' - Bn"W
(I' = 1+ 1, /" = I, ~I = + 1), (S.19R)
vp = vo(n' - nil) - (Bn' + Bn,,)1 + (Bn' - Bn,,)f2
(l' = 1- 1, I" = I, ~I = -1). (S.l9P)
Equations (S.19) given excellent agreement with the empirical fine structure
of the infrared bands.
For the HCI molecule the values of Bn have been obtained for the various
bands n' +-+ nil :
0+-+2 0+-+3
(transitions with ~n > 1 occur as a consequence of the small anharmonicity;
cf. Figure 1.3). The results are summarized in Table S.l. The difference ~Bn
between successive values is very nearly a constant, so that Bn can be fitted
by the formula
Bn = Be - (Xe(n + !), (S.20)
where (Xe is a constant small compared to Be = 1O.S909 cm - 1, the equi-
librium value of Bn. The value of Bn given in the table for the rotation vibra-
tion spectrum agree within the accuracy of the measurements with the value
bHCl = 10.438 cm - 1 obtained from the pure rotation spectrum of HCl,
Equation (4.21).
Table 5.1 Rotational constants of HCl

in the different vibrational levels of the
electronic ground state. [From Herzberg
(1966), p. 800, with permission.]
n Bn (em-I) ~Bn (em-I)
0 10.4400
0.3034
10.1366 0.3037
2 9.8329 0.2986
3 9.5343 0.302
4 9.232 0.299
5 8.933
Experimental values for the vibration-rotation spectra of the CO molecule

are not as numerous or as accurate. From the spectrum depicted in Figure
1.S one obtains
(S.21)
This value, and hence the fine structure in the energy spectrum, is consider-
ably smaller than that for HCI.
The diatomic molecule with the largest rotational constant Be, and thus
the largest energy difference between rotational levels, is the H2 molecule,
for which B~2 = 60.80 cm - 1.
A qualitative theoretical explanation of (S.17) with (S.20) follows from the
classical picture of the diatomic molecule as two rigid spheres connected by
a spring. When this system is in a state of higher vibrational energy, it has a
larger amplitude and consequently a larger moment of inertia. Consequently
I;; 1 decreases with increasing n, as expressed by (S.20).
For the quantum-mechanical observables, the empirical formula (S.20)
means that in the presence of an interaction between vibrational and rota-
tional degrees of freedom, the form (S.13) is not sufficient. For the energy
operator one has in addition to (S.12) an interaction term, for which one may
try as a first guess
(S.22)
where g is a coupling constant of dimension (e V) - 1. Thus the energy operator
for the vibrating, rotating, interacting diatomic molecule in this approxima-
tion will be given by
(S.23)
where
Hose = hw(N + tJ) (S.24)
and
(S.2S)
Ie is the moment of inertia that corresponds to the equilibrium separation

xe:Ie = J1X;.
The energy values of the diatomic molecule with vibration-rotation
interaction are the expectation values of H in the physical states. It is, of
course, not obvious that the In 1m) = <Pn ® f~ of (S.11), where <Pn are eigen-
vectors of N and f~ are eigenvectors of L2 and L 3 , represent the pure states
of this physical system. However, as they happen to be also eigenstates of the
energy operator H of (S.23), they are the obvious choice for states in an energy
measurement. Thus the energy values are the eigenvalues of H in the basis
In 1m) given by (S.l1) which are calculated to be:
1 h2
Enl = hw(n + t) + - h 2 1(l + 1) + ghw - (n + t)l(l + 1). (S.26)
2Ie 2Ie
The wave number of the radiation quantum corresponding to the energy

value, i.e.,
Enl Enl (S 27)
Vnl = hZ = 2nhc .
is called the term value 10 [cf. Equation (1.12)]. The term values of the vibrat-
ing rotator are therefore, by (S.26),
Vnl = vo(n + !) + (Be - ain + !»l(l + 1), (S.28)
where we have used the standard notation of molecular spectroscopy:
B _ h _ h
(S.29)
e - 4ncI e - 4ncJ.lx; ,
gvo.h 2
a = --- (S.30)
e 2Ie '
(S.31)
According to the above qualitative considerations I; 1 and therefore Bn
decrease with n, thus ae should be larger than zero, which is experimentally
always fulfilled. Equation (S.28) with (S.31) gives for the wave numbers of
the transitions in the R branch
= vo(n' - nil) + 2B n, + (3Bn' - Bn,,)l + (Bn' - Bn,,)12, (S.32R)

and for the wave numbers in the P branch
(S.32P)
i.e., the empirically well-established formulas (S.19). This also shows that our
first guess (S.22) was a good guess.
Two previously mentioned effects have not been taken into account in
(S.23) with (S.24) and (S.2S). These are the anharmonicity of the oscillator
and the influence of centrifugal forces. Thus (S.28) is not the end of the story
of the vibrating and rotating diatomic molecule. If these effects are also taken
into account, then up to a certain degree of accuracy one obtains for the
term values of a vibrating rotator
Vnl = we(n + !) - we~e(n + !? + Bnl(l + 1) - Dn12(l + 1)2, (S.33)
where ~e is a small parameter expressing the anharmonicity,11 and
Bn = Be - ae(n + !), (S.34)
Dn = De + f3e(n + !). (S.3S)
10 Note that we use the same symbol v for the frequency (in sec-I) and the wave number (in
cm- I); cf. statement following Equation (1.12c).

11 we~e is given by the anharmonicity of the oscillator (terms proportional to Q3, Q4, etc.)
and can be calculated by perturbation theory. See Problem VIlLI.
We is the standard notation for
We = 2:c 2:cjf·
=
(5.36)
Be is given by (5.29):
h h
B =--=-- (5.29)
e 4ncI e 4nc/1x; .
According to the semiclassical consideration leading to (4.20), De may be

expressed in terms of the reduced mass /1, equilibrium separation Xe and
spring constant k as
h3
(5.37)
De = 4nck/12x~·
From (5.36), (5.29), and (5.37) it follows that the three system parameters
De, Be, We are not independent but are related by
(5.38)
The parameters ~e, iY. e, f3e, expressing the degree of anharmonicity, are known
empirically to be small:
iY.e f3e ~ 1
~e ~ 1, Be "'" 1,
-~
De '
which must be the case, as they represent the effect of corrections to models
which are very well realized by microphysical systems in nature. The system
parameters We' ~e, Be, De, iY. e, and f3e have been experimentally determined
for many diatomic molecules and are collected in tables [cf. Herzberg
(1966)]. Equation (5.33) gives a very good description of the vibration-
rotation spectra of diatomic molecules, and only in exceptional cases are
higher correction terms needed.
Diatomic molecules are vibrating rotators only as long as the internal
energy is sufficiently low-roughly, in the region of energy of infrared
radiation. For higher energies (1-20 eV) which correspond to the visible
and ultraviolet regions, the molecules are no longer just vibrating rotators,
since new degrees of freedom become accessible to electronic transitions.
In each electronic state the molecule is, however, still a vibrator, in the same
way as in each vibrational state the molecule is a rotator. This leads to energy
spectra as depicted schematically in Figure 5.3 for two electronic states. We
shall not describe the electronic structure of molecules here, they are of the
same nature as the electronic structure of atoms which will be discussed in
later chapters of the book.
Problems 155
5
-0-----------
B
0
--------~==~I~O---- ~
r
- - 1= s0 _ _ 3
0--
r
10 = s
r --
--
0-----
-
----~
- 10 -
..·O-------
5
-s
~O-----------
A
0
Figure 5.3 Vibrational and rotational levels of two electronic states A and B of a mol-
ecule (schematic). Only the first few rotational and vibrational levels are drawn in each
case. [From Herzberg (1966), with permission.]
Problems
1. Let Pi and \'j (i,j = 1, 2, 3) satisfy the canonical communication relation [P;, Qj] =
(hji)bijI. Define the orbital angular momentum Li == (ijkQjP k (L = Q x P). The
component operators are Hermitian [Equation (2.11a)] and satisfy the commutation
relation (2.14). Define H 3, H +, and H _ as follows:
(a) Show that
and that
Show that
L2 == LiLi = h2(H+H_ + H~ - H 3)
= h2(H_H+ + m+ H3)'
(b) Show that H2 == h- 2 L 2 is an invariant operator of 0"(SU(2)); i.e., show that
and
and consequently that

[UZ, A] = 0,
where A is any element of the algebra generated by the L;,
(a, ai, aij, ... E C).

(c) Prove that the spectrum ofL 2 is h2 l(l + I) (I = 0, 1,2,3, ... ). Hint: Express the
operator L3 in terms of the annihilation operators ai and creation operators a;:
1
ai = h>.. (Qi + iPi),
v' 2h
1
at = J2h (Qi - iPJ
(units chosen so that )1W = I) of the three-dimensional harmonic oscillator, and

show that as a consequence of the spectrum of the harmonic oscillator L3 can
have only integral eigenvalues.
2. Calculate the internuclear distance for the CO molecule using the absorption
spectrum given in Figure 1.7.
°
3. The energy-loss spectrum of vibrating H2 molecules (Figure PS.l) shows two
bumps at and at 0.52 eV with respective intensities 3.5 and 7.8 x lo = 0.26 (arbitrary
units), respectively. What is the statistical operator W for the ensemble of H2 molecules
I
Sweep Voltage, volt
Figure PS.l Energy-loss spectrum of H2 [From G. 1. Schultz, Phys. Rev. 135, A988
(1964) with permission.]
Problems 157
in this energy-loss experiment? Can dipole transitions occur in this Hz gas? What
would be the frequency for these dipole transitions?
4. In an infrared absorption experiment with HCI molecules in the ground state it is
observed that the following frequencies v( cm - 1) are absorbed:
20.68 82.72 144.76 206.80
41.36 103.40 165.44 227.48
62.04 124.08 186.12
Assume that in this absorption experiment only transitions between neighboring energy
levels take place (dipole transitions). What are the energy levels of this (rotating) HCl
molecule if the zero of the energy scale is fixed by Eo = O?
5. Let {fj-~lW,f{iz1IZ} be the orthonormal basis for the space ~lIZ, defined in Section
1II.3.
(a) Show that the matrices (Ji with matrix elements (Ji = (2/h)(fi,.,~ liZ, JJ~~ liZ) are
given by
with respect to the basis {f~~lW, f{i/ IZ }. The (Ji are the Pauli spin matrices.
(b) Show that the Pauli spin matrices together with the unit matrix are a basis for
the vector space of operators on ~lIZ.
6. Let and ~ 2 be two finite-dimensional linear scalar-product spaces and ~ =
~1
~ 1 (B ~ Z be their direct sum. Let A 1 be a linear operator in ~ 1 and A z be a linear opera-
tor in ~ 2' The direct sum Al (B A2 is defined as the map
(AI (B A 2)1/I = (AI (B A 2)(1/I1 (B I/Iz) = (Ali/ll EB Azl/l z)
for I/Ii E '~i and 1/1 = 1/11 (B I/Iz E~. Let'yf = 9f 1 ® 9f 2 be the direct product of 9f 1
and [llz. The direct product Al ® A z is defined as the map
(AI ® A 2)h = (AI ® A z)(h 1 ® h2) = A1hl ® A2h2
for hi E [lli and h = hi ® h z E:if.
(a) Show that in an appropriate basis the matrix a of the operator Al (B A2 has the
form
( a1
o
0),
a2
where al is the matrix of Al and a2 is the matrix of A 2.

(b) Show that the matrix of the operator Al ® A z when written in an appropriate
basis has the property
i.e., may be written as a block matrix

aWa(z)
dIn1 )a(2))
( aWa(2) a~lJa(2)
a~11)a(2) '" a~~)a(Z)

or
a(Z)a(1)
11
... dZ)dl))
1m
( ·· ..
· .
a~la(1) a~~a(l)
where di) is the matrix of Ai (i = 1, 2) and a~~ are the matrix elements.
(c) Show that
(i) Tr(AI EB A z) = Tr Al + Tr A z ,
(ii) Tr(AI ® A z) = Tr Al Tr A z .
7. Let A, B, and C be three linear operators which fulfill [A, BJ = C. Under what
condition can a vector fbe simultaneously an eigenvector of A and an eigenvectorof B?
CHAPTER IV
Complete Systems of Commuting

Observables
In this chapter it is explained that the question of what constitutes a complete

system of commuting observables is not a mathematical question but can
only be answered by experiment.
For the one-dimensional oscillator we needed one number to label the
basis vectors of the space of physical states. We used the energy (or principal)
quantum number n, which is the eigenvalue of the observable N of Equations
(11.3.2)-(11.3.7) and which is related to the energy by (11.3.30):
Nln) = nln) or Hose In) = Enln), n = 0, 1,2,.... (1.1)
Instead of the quantum number n one could have used the "continuous
quantum number" x, i.e., the generalized eigenvalue of the position operator
(11.7.1). Instead of the basis vectors In) one could have used the generalized
basis vectors Ix) which are connected with the In) by the transformation
(11.7.16'). One could as well have used the generalized basis vectors Ip) of
(11.7.28) and (11.7.37), or even other basis systems. All these basis systems
for the one-dimensional oscillator are labeled by one quantum number.
One says that the spectra of N, H, Q, and P are simple or nondegenerate. 1
For the rotator we needed two quantum numbers for which we had chosen,
in Section 111.3, the integer or half-integer numbers j and j 3' which are
connected to the eigenvalues of the square of the angular momentum opera-
tor J2 and the arbitrarily chosen component J 3 :
J 2Jj,j3) =j(j + 1)h 2 Ij,j3); J 3Jj,j3) =j3IiJj,j3)' (1.2)
I These observables are cyclic operators for the space of physical states ofthe one-dimensional
oscillator. Cf. footnote 4 of Section 1.4.
159
160 IV Complete Systems of Commuting Observables
The spectrum of J2 is not simple, it is degenerate, and in order to label the

basis vectors one needs in addition to j the quantum number j3 (the "mag-
netic" quantum number).2
Instead of the basis vectors (1.2) one could use another system of basis
vectors for the space of physical states [JIl of the rotator, e.g., the ij j2} defined
by
(1.3)
Equations (1.2) or (1.3), along with normalization, completely define the
basis vectors up to a phase factor (i.e., the subspace spanned by the basis
vectors). This means: If J21/1 = al/l and 131/1 = bl/l for 1/1 E [JIl, then a =
j'(j' + 1)112 where j' is an integer or half-integer, b = j3 where j3 is an integer
or half-integer with - j' s;, j3 s;, j', and 1/1 = (1.1 j', j3) with rt. E C. The two
basis systems (1.2) and (1.3) are related by a basis transformation, (1.5.12),
with transformation coefficients (j j31j' j~} which are completely determined
by the algebra of observables, except for phase factors (which are usually
fixed by a convention).
Whereas for the oscillator one needs one observable and one quantum
number related to it (e.g., Hose and n), for the rotator one needs always two
commuting observables and two quantum numbers related to them (e.g.,
J2, 13 andj,j3)' To a certain extent one has a choice of which observables one
wants to take, but the number of labels needed to specify the basis vectors of
the space of physical states is dictated by the physical system. 3
A system of commuting Hermitian operators that specifies the (general-
ized) basis system completely is called (following Dirac) a complete system
of commuting observables (c.s.c.o.). The (generalized) eigenvalues of a C.S.C.O.
are called quantum numbers.
For the one-dimensional harmonic oscillator, the C.S.C.O. consists of one
operator, either the operator H, or the operator Q, or some other operator.
For the rotator the C.S.C.O. consists of two operators; J2 and 13 are a conven-
ient choice, but others are also possible, even operators with continuous
spectrum. 4 For different physical systems and, therewith, different algebras
of observables one usually has different c.s.c.o.'s. The larger the algebra
(i.e., the more complicated the physical system), the larger is the number of
operators in a C.S.C.O.
From what is said above and the fact that in a direct-product space a basis
system is obtained as the direct product of the basis vectors in the two factor
2 None of the observables J i or J2 are cyclic operators in the space of physical states of the
rotator.
3 It can happen that in the transition from a discrete quantum number to a continuous
quantum number a new label-which usually takes a finite number of values-will appear so
that the set of discrete quantum numbers for a given physical system may contain fewer quantum
numbers than the set containing one or more continuous quantum numbers. Thus the number
of observables for the complete systems of commuting observables of a given physical system
is not necessarily a fixed number.
4 Such a basis system is given in (3.76) of the Appendix to Section V.3.
IV Complete Systems of Commuting Observables 161
spaces, it is clear that one c.s.c.o. of the combination of two systems is given
by the combination of the two c.s.c.o.'s for each subsystem.
For example, a c.s.c.o. for the vibrating rotator is given by
(1.4)
But there are other possible c.s.c.o.'s for the vibrating rotator. Some of
them may gain physical importance when the interaction between rotator
and vibrator becomes strong and the assumption that the Hamiltonian
of the combined system commutes with (1.4) becomes a poor approximation.
[The problem of determining for which operator *-algebras there

exists a C.S.C.o. is unsolved; the requirement of the existence of a
C.S.C.o. is certainly a restrictive condition. For certain classes of oper-
ator *-algebras.9I there do exist c.s.c.o.s (e.g., if .91 is the enveloping
algebra of a nilpotent or semisimple group).]
The physicist's problem is not to find a c.s.c.o. for a given algebra, but is
usually the reverse: From the experimental data he finds out how many
quantum numbers are required, and what are the possible values of these
quantum numbers. This gives him, according to the basic assumption IlIa,
the complete commuting system {Ak} and its spectrum. He then conjectures
the total algebra .91 by adding to {Ad a minimum of other operators such
that the matrix elements of elements of .91 calculated from the properties of
this algebra agree with the experimental values of the corresponding ob-
servables.
Thus, the question of what is a c.s.c.o. for a particular physical system and
the question of when a system of commuting operators is complete are
physical questions. If an experiment gives more values than can be supplied
by a given system of commuting operators, then this system is not complete;
one has to introduce a new quantum number, i.e., enlarge the system of
commuting operators. This usually requires a further enlargement of the
algebra.
An example of this procedure is the transition from the rotator model of
the diatomic molecule to the vibrating rotator model: As long as the energy
is small (;$1O- z eV) only rotational states can be excited. The diatomic
molecule is described by the algebra of operators of the rotator and the
quantum numbers are j and j 3 ; J Z , J 3 is a c.s.c.o. If the energy is increased
one observes -new levels, each of which are the starting points of a whole
rotational band, as depicted in Figure 5.1 of Chapter III. To describe these
levels one introduces a new quantum number n and assigns the values 0, 1,2,
etc. of this new quantum number to them. With the new quantum number
n one has introduced a new observable N so that the C.S.C.o. is now (1.4).
Then one adds to this set of operators further operators at and a which
together with N form the algebra of observables such that the other features
of the physical system (e.g., the equal spacing between the levels) are correctly
described.
162 IV Complete Systems of Commuting Observables
Not only can an increase in energy, i.e., an extension of the domain of

applicability of the theory, require an enlargement of the c.s.c.o. and of the
algebra of observables, but so also can an increase in the precision of measure-
ment. This can also be illustrated by the vibrating rotator model:
As long as one could not resolve the fine structure of the order of 10- 3 eV,
the quantum number n was sufficient to describe the infrared spectrum of the
diatomic molecules and the oscillator model provided an approximate de-
scription which was accurate enough for coarse-resolution experiments
(every theoretical description is an approximate description). However,
when the results of high-resolution experiments had to be explained, new
quantum numbersj andj3 (to explain splitting in a magnetic field) had to be
introduced. The c.s.c.o. J2, J 3 corresponding to these new quantum numbers
is then augmented by the transition operators to form the whole new rotator
algebra which is then combined with the oscillator algebra to give the algebra
of observables of the vibrating rotator.
Because of the above described procedure for obtaining the algebra of
observables d for a given physical system, we can assume that for the algebras
in quantum mechanics one always has a C.S.C.o. Thus for the algebra d there
exists a set of commuting operators
A l ,A 2,···,A N Ed (1.5)
which have a set of (generalized) eigenvectors,s
AkIAl"'" AN) = AkIAl"'" AN), (1.6)
<fJ r
such that every physical state vector <fJ can be represented as
= dll(A)IAl···AN)<Al···ANI<fJ)·
Here the set A = {(AlA2 ... AN)} is the spectrum of the c.s.c.o. (1.5) with
(1.7)
Ak = {Ad being the spectrum of the observable A k. This means that Ak is a

set of generalized eigenvalues if Ak has a continuous spectrum, it is the set
of all discrete eigenvalues if Ak has a discrete spectrum, and it is the set of all
discrete and continuous eigenvalues if Ak has a discrete and continuous spec-
trum. SA dll(A) = f"" SA2 ... SAN dll()'l"'" AN) means summation over the dis-
crete spectrum and integration over the continuous spectrum. Thus (1.7)
is the generalization of the spectral theorem (1.4.4g) from the one variable
a to the N variables AI' ... , AN'
It often happens that some of the observables A l A 2 , •.. , AM of the
c.s.c.o. have only a (absolutely6) continuous spectrum Ac = {(AI' ... , AM)}
and the other observables AM+ 1 " ' " AN of the c.s.c.o. have only a discrete
spectrum AD = {(A M+ l ' ... , AN)}' Then (1.7) takes the form
¢= f···
Ac
fdll(Al,A2, ... ,AM)L···LIA1, ... ,AM,AM+l, ... ,AN)
AD
X <AI"'" ANI<fJ), (1.8)
5 Mathematically this assumption can be stated: the conditions of the nuclear spectral
theorem are fulfilled.
6 Cf. footnote 7, Section 1.4.
IV Complete Systems of Commuting Observables 163
where the summation runs over all discrete eigenvalues P'k} k = M + 1,

M + 2, ... , N of all observables AM+ P ... , AN' and the integration extends
over a set of generalized eigenvalues {A k } k = 1,2, ... , M of all observables
AI' ... ,A M •
The statement (1.7) or (1.8) is the basis of the Dirac formulation of quantum
mechanics. 7
An immediate consequence of the statement (1.6), (1.7), or (1.8) is the
following statement: Let AI' ... , AN be a C.S.C.o. with (generalized) eigen-
vectors 1,.1.1 ... AN) = 1,.1.) such that
(k=I,2, ... ,N).
Then from
(i = 1, 2, ... , N) (1.9)
it follows that
ai = Ai and la} = 0:1,.1.), (LlO)
where 0: E C.
For the case that the spectrum A is continuous, i.e., that one of the Ai can
take continuous values, the formulation of (LlO) has to be given a precise
meaning. In the following we shall use (LlO) only for the case of a discrete
spectrum.
7 It is not a theorem, because the mathematical premises characterizing the algebra d cannot
be stated. But it is based on the general nuclear spectral theorem, which is a generalization ofthe
nuclear spectral theorem for a cyclic operator discussed in Section 1.4 to a strongly commuting
family of operators A l' . . . , AN' Note that, already for the enveloping algebra of a unitary
irreducible representation of a noncompact group, the number N is not fixed.
CHAPTER V
Addition of Angular Momenta-

The Wigner-Eckart Theorem
In Section V.I the elementary rotator is defined as the system described by

an irreducible representation of the algebra of angular momentum. Section
V.2 discusses then the direct product of two irreducible representations of
angular momentum and its reduction with respect to the total angular
momentum. The Clebsch-Gordan coefficients are introduced, their recursion
relations are derived, and their most frequently used values are tabulated.
In Section V.3 tensor operators are introduced and the Wigner-Eckart
theorem for the rotation group is stated without derivation. In Section VA
a new observable, parity, is introduced. Parity is then applied to discuss the
spectrum of diatomic symmetric-top molecules. In an Appendix to Section
V.3 the irreducible representations of the algebras of SO(3, 1), SO(4), and
E(3) are derived.
V.l Introduction-The Elementary Rotator

In Chapter III we have treated the quantum-mechanical rotator. Its space
of physical states is given by (111.4.2):
(1.1)
where ~l is the irreducible-representation space of the algebra of angular

momentum of the dumbbell, L i , with
(1.2)
164
6"(SO(3)d is a subalgebra of the algebra of observables for the rotator. 1

Let us now consider the physical system whose space of physical states
is ~/, with an arbitrary fixed value I = 0,1, 1, ~, ... , and whose algebra of
observables is the angular momentum algebra. Every rotator can, under
certain conditions, be considered as such a physical system-namely, if the
energy E"xt that can be transferred to or from this rotator is small compared
to the energy differences E/ - E/- 1 and E/+ 1 - E/. In nature, this condition
is rarely fulfilled, because a state of higher energy always tends to decay into
a state of lower energy" spontaneously," and therefore the excited state has
only a limited lifetime, unless E/ is the ground-state energy value. It is,
nevertheless, customary to consider these unstable states often as independent
physical systems, not only in nonrelativistic quantum mechanics, but also
in the relativistic case of elementary-particle physics. We call a physical
system whose space of states is ~/ an elementary rotator. Elementary rotators
are not just substructures of the diatomic molecule; they appear in other
quantum-mechanical systems in all areas of physics.
V.2 Combination of Elementary Rotators
According to the basic assumption IV, the space of physical states of the
combination of two elementary rotators with spaces ~{b and ~{i) is the
direct-product space ~{h ® ~{i) = f3f. Let J~I) and J~2) be the angular-
momentum operators in ~{b and f3f{~), respectively, and let
and
be the basis systems in ~{b and ~{~)' respectively. Then a basis system in
allh IV\ alli2' .
,'71;(1) ~ ,'71;(2) IS gIven
by
Ulm l ) ® U2 m 2) = Ul m lj2 m 2)' (2.1)
The J~a) (et = 1, 2) fulfill the commutation relations

[ J~a)
I ,
j<.a)]
J
= ihf IJk
.. J(a)
k • (2.2)
We define operators in ~ = ~{b ® ~{~) by
Ji = Jj1) ® I + I ® J~2). (2.3)
1 Generators of &'(SO(3» can represent various angular momenta. For example, they may be
the orbital angular momenta of a mass point (111.2.10), or they may be the angular momenta of
a dumbbell, or they may be the spin operator of an electron, etc. That is, the generators of
&'(SO(3» fulfilling (1.2) may have various different physical realizations. Though all &'(SO(3» are
mathematically equivalent, their generators may represent different physical observables. In
order to specify which physical quantity the &'(SO(3» represents, we write the observables as
subscripts to the symbol for the algebra, e.g., as &'(SO(3)d.
166 V Addition of Angular Momenta-The Wigner-Eckart Theorem
As a consequence of(2.2) and (2.3) it follows that the J i fulfill the commutation
relation
(2.4)
Thus the J i of (2.3) fulfill the defining relations of the generators of the
algebra of angular momentum, and are thus a representation of angular
momentum in the space ~{b ® 8£{~). The J i are called the total-angular-
momentum operators of the combined system.
The basis system (2.1) of ~ consists of eigenvectors of the following
complete set of commuting operators (c.s.c.o.):
J(1)2 ® I, (2.5)
with the eigenvalues
J(1)2 ® Ilj1m1j2m2) = jl(jl + 1)h2U1md2m2),

J~l) ® IU1md2m2) = m 1hUlmlj2 m2),
(2.6)
I ® J(2)2Ij1m1j2m2) = jij2 + 1)h 2 11 1 m 1h m2)'
I ® J~2)ljlm112m2) = m2hUlm112m2)'
We now introduce the abbreviated notation for the operators in 8£:
(2.7)
These J~a) are operators in the space .qJt = ~{b ® ~{~) in constrast to the
original J~a), which were operators in 8£l;) only. The J~a) of (2.7) clearly fulfill
the commutation relation (2.2), so we denote them by the same symbol.
However, we should keep in mind that the old J~a) are the restrictions of the
new J!a) , to the space 8£i.
(a)'
The basis (2.1) is in general not a physical basis for the combined system.
A basis is physical if it consists of eigenstates in which the physical system
can be prepared. If all physical states are eigenstates of the energy operator
H of the physical system, then the physical basis must consist of eigenvectors
of operators that commute with the energy operator. In general not all the
Jl a ) commute with the energy operator (in particular, not if there is an inter-
action term between the two angular momenta).
The operators that in general commute with the energy operator are the
total angular momenta J i (they commute with H whenever H is rotationally
invariant). Therefore it is convenient to choose a basis in ~ that consists
of eigenvectors of the system of commuting operators
J(1)2 = J(1)2 ® I, J 3' (2.8)
We denote this basis by
(2.9)
It has the properties

J(~)2Ijlj2j m) = jij~ + l)h 2ljlj2j m),
J 3ljlhj m) = mhljd2j m), (2.10)
J 21jlj2j m) = j(j + l)h 2ljlj2j m).
Each vector, and consequently each basis vector of the basis (2.1), can be
expanded according to (1.4.4d) with respect to the basis system (2.9):
Ijlmlj2 m2) = L Ud2j m) (jlhj mljl mlh m 2);

j,m
(2.11)
and each basis vector of the basis (2.9) can be expanded with respect to the
basis system (2.1),
(2.12)
The transition coefficients (jlm1hm2Ijlj2jm) are called the Clebsch-

Gordan or Wigner coefficients, and are denoted by
<j1md2m2ljlhj m) = <jlmd2m21j m)
= C(jlhj m 1 m2 m)
(2.13)
= <m 1m21j m)
for fixed values of j 1 and h. In (2.13) we have given various notations used in
the literature.
Taking the scalar product of (2.11) with Ij 1 m'd2 m~) we obtain the orthogo-
nality relation of the Clebsch-Gordon coefficients:
6 m',m,6m'2m2 = I (jlm'lhm~lj m)<j mljlm 1j2 m2), (2.14)

i.m
and taking the scalar product of (2.12) with Ijljd m'), we obtain
6jj ,6 mm , = L <j' m'1j 1md2 m2)<j 1md2 m21j m). (2.15)
We now want to find the spectrum of J2 and J 3 in fJ£ = fJ£{b @ fJ£{~),

i.e., the values that j and m in basis system (2.9) can take.
For the sake of clarity in the following discussion we use curly brackets
for the direct product basis vectors,
to distinguish them from the total angular momentum basis vectors Ijlj2j m).
Also, since j 1 and j 2 are fixed values, we just write
m.=
-3 -2 - I 0 2 3
Figure 2.1 Dot pattern of direct product basis vectors Im1m Z } = U1m l >® Ijzmz> in
the space f1l = f1l{lj 3 ® f1l{ij 3/Z. Shading indicates those basis vectors for which
Iml2 U, - hi·
Since the vectors Im 1 m2} are eigenvectors of J 3 = J~l) + J~2) with eigen-
value ml + m2, the spectrum of J 3 in fJ£ is given as all possible sums of
m 1 =jl,jl - 1, ... , -jl withm2 =h,j2 - 1, ... , -j2:
(2.16)
Since the vectors Ij m) are eigenvectors of J 3 with eigenvalue m, the quantum

number m in the basis Ij m) also ranges over the values (2.16). The subspace
of fJ£ consisting of all the eigenvectors of J 3 with a particular value for m is
denoted by fJ£m' The dimension of fJ£m is obtained by counting the number of
direct product basis vectors Iml m2} for which m 1 + m2 = m and it is given by
for Iml ~ Ul - j21, (2.17)

for Iml ::; Ijl - hi·
Figure 2.1 illustrates the case j 1 = 3, h = ~ and suggests how this counting
is done for general j 1 and j 2' The dimension of a space fJ£m is obtained by
counting the number of dots along the diagonal for which m 1 + m2 = m.
The space fJ£ h + h is, according to (2.17), one-dimensional and it is spanned
by the direct product basis vector Udz}. The vector Ulj2} is also an eigen-
vector of J2 with eigenvalue (jl + h)(jl + j2 + 1):
PROOF. We have in general
J2 = (J(l) + J(2))2 = J(1)2 + 2J(l ) • J(2) + J(2)Z

= J(l)2 + J(2)2 + 2(J\')J\2)+ J~I)J~2) + J~1)J~2))
= J(I)2 + J(2)2 + 2J~I)J~2) + J<!:)J<;) + J<.!.)J<;). (2.18)
Now,
(2.19)
since j 1 and j 2 are the highest possible values of m 1 and m2, i.e., since J~)!j Ij 1> = 0
and J<;)!j2j2> = O. Thus
J 2Ulj2} = (J(1)Z + J(Z)Z + 2J~l)J~Z)Uljl> ® U2j2>
= [jl(jl + 1) + h(j2 + 1) + 2jlj2]fl2[jljl> ® U2h>
= (jl + jZ)(jl + j2 + lWUlj2}. 0 (2.20)
As Ulj2} is an eigenvector of the c.s.c.o. (2.8), it is related by a complex number to
the total angular momentum basis vector U1 + h j 1 + j 2> which has the same eigen-
values and spans the same one-dimensional subspace of [Jl:
exEC. (2.21)
Assuming both basis vectors are normalized, ex is a phase factor and we may choose
it to be unity so that
(2.22)
and
(2.23)
Besides spanning [Jljt + J2 , UI + j2 jl + j2> belongs to the space [jfh+ h of all eigen-
vectors of J2 with eigenvaluej(j + 1) = (jl + j2)(j1 + j2 + 1). Any such eigenspace
[jfJ of the total angular momentum is invariant under the action of J; = J)I) + J\2)
and has the properties of an elementary rotator space with respect to these total
angular momentum operators. Thus, since [jl + j2 jl + j2> has the highest value
for m of any vector in [jfh + h, a basis of total angular momentum vectors !j m>
may be obtained in [Jlh + h by applying
J _ = Jf.!.) + J<!) (2.24)
successively to [jl + j2 jl + j2> following the procedure of Section III.3. One obtains
by this procedure the 2(j 1 + h) + 1 basis vectors
UI + j2 m>, (2.25)
Now consider the space [Jlh + J,- 1 of eigenvectors of J 3 with eigenvalue m = jl +
j2 - 1. According to (2.17), this space is two-dimensional (unlessjl = 0 or j2 = 0).
One total angular momentum basis vector in [jf h + iz _ 1 is the vector UI + j 2 j I +
j2 - 1> which already appears in (2.25) as a basis vector in [jfh + iz. Due to the proper-
ties of the spaces [Jli, the other basis vector U m = jl + j2 - 1> in ~h + i,- 1 must
belong to a space [JlJ for which j = jl + j2 - 1 + n, n = nonnegative integer.
However, it certainly cannot belong to a space [jfi for which j > j 1 + j2 since this
would imply that values of m larger than j 1 + j2 occur in PA and it also cannot belong
to the space ~h + 12 since it would then not be orthogonal to [jl + j2jl + jz>. Thus,
this new vector belongs to the space [jfh + i,- I. This determines the basis vector
[jl + j2 - Ijl + j2 - 1> up to an arbitrary phase which is fixed by requiring
Ulj2 - lUI + j2 - 1 jl + j2 - 1> be real and positive. (2.26)
As in the case of [Jlh+12, since [jl + j2 - 1jl + j2 - 1> has the highest value for
m of any vector in [Jlh + h - I, a basis in [Jlh + 12 - I is generated by applying J _ succes-
sively to UI + j2 - 1 jl + j2 - 1> to give the 2(j1 + j2 - 1) + 1 basis vectors
[jl + jz - 1 m>, m = jl + h - 1,jl + j2 - 2, ... , -(jl + h - 1).
(2.27)
170 V Addition of Angular Momenta ~ The Wigner-Eckart Theorem
According to (2.17), the next J 3 eigenspace, ~h + h- 2, is three-dimensional (unless

j, = 0, t or h = 0, t)· We have already found in ~jl+h-2 the orthogonal vectors
Ij, + j2j, + h - 2) from (2.25) and Ul + j2 - 1 jl + j2 - 2) from (2.27) and need
one more total angular momentum basis vector orthogonal to these to span ~ h + h - 2.
By an argument similar to that given in the previous paragraph, the new basis vector
U m = j, + h - 2) must have j = jl + j2 - 2 and so belongs to the total angular
momentum space ~h + h-2 spanned by the 2(jl + j2 - 2) - 1 basis vectors
Ijl +h - 2 m), m = jl +h - 2,jl +h - 3, ... , -(jl + j2 - 2).
(2.28)
The phases of these vectors are fixed by requiring
{jlh - 2U, + j2 - 2jl + j2 - 2) be real and positive. (2.29)
We continue in this way and consider in succession the J 3 eigenspaces ~m' m =
jl + j2' j, + j2 - 1, ... , U, - j21. At each step a new total angular momentum
space ~j=m must be introduced which contains the vector 1m m) needed to complete,
along with previously determined vectors
Ul + j2 m), Ul + j2 - 1 m), ... , 1m + 1 m),

the basis for the space ~m' This is so because, by (2.17), the dimensionality of the
successive spaces ~m increase by one until ~Ih _ hi is reached. Once ~Ih - hi has been
considered, and the last space ~j= Ih - 121 introduced, none of the remaining J 3 eigen-
spaces ~m' m = Ul + j21 - 1, Ul - j21 - 2, ... , -(jl + j2), have dimension greater
than that of ~Ih _ 121 and they are in fact spanned by total angular momentum basis
vectors already found in the spaces ~jl + 12, ~h + h- 1 , ••• , ~Ih - hi. Thus the values
j may take are j = jl + h, j, + jz - 1, ... , U, - jzl. The phase choices (2.23),
(2.26), (2.29), etc. are summarized in the convention
This is the standard convention for the Clebsch-Gordan coefficients.

Therewith we have seen that the spectrum of J2 in ~ is given by
spectrum J2 = {j(j + 1): j = jl + j2,jl + h - 1, ... , Ijl - j2i}, (2.31)
and the space ~ may be expressed as
~ == ~{h ® ~H) = ~h+h EB ~h+jz-' EB··· EB ~h-h. (2.32)
This is called the reduction of ~ into a sum of irreducible total angular momentum
spaces.
Summarizing, we have seen that the space ~{h ® ~H) is in general not an irre-
ducible-representation space or ladder-representation space of the algebra of total
angular momentum 6'(SO(3)J). That is, not all vectors of ~{h ® ~{~) can be obtained
by applying the J ± a sufficient number of times to one of its vectors, but rather it is the
direct sum of several such irreducible representation spaces ~j as given in (2.32). We
have seen in particular that the Clebsch-Gordan coefficients, which are the transition
coefficients between the two basis systems (2.1) and (2.9), are zero unless j is as in
(2.30) and m = m, + m 2 :
{m,m2U m) = °
j # j, + h,j, + j2 - 1, ... , Ih - M (2.33)
The Clebsch-Gordan coefficients are calculated recursively.2 The recursion

relations are obtained if one takes the matrix element of J ± = J~) + J<1' between the
states Ij m) and Iml m2}
{mlm2lJ ± Ij m) = jj(j + 1) - m(m ± 1)1i{mlm2Ij m ± I)

= {mlm2IJ~) + J<f)lj m)
= «J~) + J\f)lmlm2}, Ij m»
= Jjl(jl + 1) - ml(m l =+ l)fi{ml =+ 1 m21j m)
+ jh(j2 + 1) - mim2 =+ 1)1i{mlm2 =+ Ilj m). (2.34±)
Here we have used (111.3.21-23),
L± J~ = jl(l + 1) - m(m ± I)1iJ~± I
for
and
In the first step we shall describe the calculation of the Clebsch-Gordan coefficients
{mlm2lj j); in the second step we then turn to the calculation of Clebsch-Gordan
coefficients for all values of m, {mlm21j m).
Setting m = j in (2.34 + ), we obtain
-jUjl + 1) - ml(m l - 1){ml - 1 m21j j)
=jj2(j2 + 1) - m2(m2 - l){mlm2 - IIj j). (2.35)
Because of (2.33) we can restrict our attention to the values ml + (m2 - 1) =

(ml - 1) + m2 = j( =m). Then (2.35) can be written
{m - 1 . - m + 11 ..) = _ [(j2 + j - ml + 1)(h - j + m l )]1/2 {m . - mi· .)

I } I 1J (.h + ml )( h. - ml + 1) I} I }}
(2.36)
Starting from ml = jl one can calculate all {mlm2lj j) successively from (2.36). One
obtains for an arbitrary ml and m2 = j - ml
{mlm2Ijj) = (_I)h -m,
x [(j2 + j - jl + 1)(j2 + j - jl + 2)··· (j2 + j - m1)(j2 - j + jl)

2jl(2jl - 1)··· (2iJ - jl + ml + 1)
(j2 - j + jl - 1)··· (h - j + ml + 1)]1/2 {.. . I ..)
x 12.... (h .-m) h}-h1J
l[
l
=(_I)h- m (j1+ml)!(j2-j+jl)!(j2+j-ml)! 1/2

(2jl)! (jl - ml )! (j2 - j + ml)! (il + j - jl)!
X {jlj - jlljj). (2.37)
2 In the following a brief outline of the calculation of the Clebsch-Gordan coefficients is

given, only the results given in the table will be needed later and this part may, therefore, be
omitted at a first reading.
172 V Addition of Angular Momenta - The Wigner-Eckart Theorem
From the orthogonality relation (2.15), it follows that for j = j' = m = m'
+h
1 = I l{m 1 m2Ijj)1 2. (2.15')
Inserting (2.37) into (2.15') and making use of the equality 3
I (jl + m 1 )! (j2 + j - ml)! = (jl + j2 + j + I)! (-jl + j2 + j)! (jl - j2 + j)!

m, (jl - ml)! (j2 - j + m 1 )! (2j + I)! (jl + j2 - j)! '
(2.38)
one obtains
•• • •• 2 (2jl)! (2j + I)! )
lUI] -hl}})1 = ('h +.12]
+. + 1)'·('h _.12 + ].
')' (2.39
And with the phase convention (2.30) one obtains

.. ... (2jl)!(2j+l)!
Ul] -hi}}) = + (.h +'12]
+' + 1)'.(.] _ 12. + ].
')' (2.40)
This equation, together with (2.37), gives all coefficients that occur on the right-hand
side of Equation (2.34) for m = j. Equation (2.34-) then gives the values of
{mlm2Um=j-l)
and successively the values of {mlm2U m) for all m = j - l,j - 2, ... , -j.
In this way a general formula for the Clebsch-Gordan coefficients can be derived. 4
The result of the above described derivation is the following formula for
the Clebsch-Gordon coefficients:
<j1md2m2U m) = {m1m2U m)
=b [(2j + 1)(j1 + jz - j)! (j1 - jz + j)!( -j1 + jz + j)!J1 /Z
ml+ m2,m (j1 + jz + j + I)!
X [(j1 + m 1)! (j1 - m 1)! (j2 + m2)! (jz - m2)! (j + m)! (j - m)!r IZ
x L {( -1Y/[z! (j1 + j2 - j - z)! (j1 - m 1 - z)! (jz + mz - z)!
z
x (j - j2 + m 1 + z)! (j - j1 - m2 + z)!]} (2.41)

The summation index z ranges over all integral values for which the argument
of every factorial involving z is nonnegative. In (2.41) we have returned to the
notation (2.1) for the direct product basis, but we shall still use Um) for the
total angular momentum basis.
We also list some of the properties (symmetry relations) of the Clebsch-
Gordan coefficients:
(j1 m1jz m2Ij3 m3) = (_I)h+m2[~!3: ~Jl/2 <jz - mZh m3U1 m1)'
11 (2.42)
We now give explicit expressions for the Clebsch-Gordan coefficients for
the cases jz = 0, t, 1.
3 A derivation of (2.38) can be found in Edmonds (1957), Appendix 1.

4 A derivation using the same notation as used here is given in Edmonds (1957), pp. 44-45.
Table 2.1 (jj m - m 2 t m 2 Ij m)

m2 =1 m2 =-1
j = j1 + t 11 +m +t 11 - m +1
+1
2j1 2j1 +1
. . 1 11 - m + t 11 + m + t
}=l!-z
2j1 + 1 2j1 + 1
For j2 = 0,
(2.43)
The values for j2 = ! andj2 = 1 are given in Tables 2.1 and 2.2.
Instead of the Clebsch-Gordan coefficients it is frequently more con-
venient to use the Wigner 3-j symbols, since these display symmetry prop-
erties more clearly. The 3-j symbol is defined by
( jl j2 j3)-(_1)it-h-
-
m 3(2· +1)-1/2<·
13 I· - m3 .
h mlh. m213 >
ml m2 m3
(2.44)
Its symmetry properties are given by:
Cl
m1
h j3 )
m 2 m3
= C2
m2
h
m3
jl )
m1
= C3
m3
( - l)it + h + h C 1
ml
j2
m2 m3
h) = (h
m2
jl
m1
j3 )
m3
= Cl
m1
h
m3
j2)
m2
= C3
m3 ~2 ~~), (2.46)
and
( j1 j2 h)=(_l)it+h+h( jl j2 j3) (2.47)
ml m2 m3 -ml -m2 -m3
Table 2.3 5 lists several values for the 3-j symbols from which many others
can be calculated using the symmetry properties (2.45)-(2.47)
5 From Edmonds (1957), with permission.
Table 2.2 (j1 m - m 2 1 m 2 1j m)

j= m2 =1 m2 =0 m2 = -1
(j1 + m)(j1 + m + 1) (j1 - m + 1)(j1 + m + 1) (j1 - m)(j1 - m + 1)
j1 +1 (2j1+ 1)(2j1 + 2) (2j1 + 1)(j1 + 1) (2jj + 1)(2j1 + 2)
(j1 + m)(jj - m + 1) m (jj - m)(j1 + m + 1)
j1
2jj (j1 + 1) + 1)
Jj1(jj 2j1(j1 + 1)
(j1 - m)(j1 - m + 1) (j1 - m)(jj + m) (j1 + m + 1)(j1 + m)
j1 - 1
2j1(2jl + 1) j1(2j1 + 1) 2j1(2jl + 1)
~
-<
~
Table 2.3 g:
o·
::l
h j3)=(_1)J I2[(jl+j2-j3)!(jl+j3-j2)!(h+h-jl)!J1 /2 (11)! . o
-,
( jl
000 (jl+j2+h+l)! (P-jl)!(tJ-j2)!(tJ-h)! ~
(Jq
if J is even, s::
..,sr
j3) h ' ~
I IS 0 o
0
( jlo j2 0 = O'fJ' dd , were J = 11 + 12, + 13,
8
g
( J+t 1 t) J-M-1 /2[ 1-M+t J1 /2 (J + 1, 1, t) £'
M -M-t t (-1) (21+ 2)(2J + 1) I
>-3
::r
(1)
1 + 1 J 1) J-M-l[ (J - M)(J - M + 1) J1 /2
(
M -M-11 (-1) (21+ 3)(2J+ 2)(21 + 1) ~
(J + 1, J, 1) <§'
J +1 1 (1)
( (_1)J-M-l[(J + M + 1)(1 - M + 1)'2J1 /2 ..,
I
M -M ~) (21 + 3)(2J + 2)(21 + 1) tTl
~
po
J (_V- M [(l - M)(J + M + 1)'2J1 /2 ::\.
(~ -M-1 ~) (21 + 2)(21 + 1)(21) >-3
(1, J, 1) ::r
(1)
1 1) ( V- M M o
..,
(1)
(~ -M 0 [(2J + 1)(J + 1)J] 1/2 8
J (_1)J-M+1/ 2[(J - M - t)(J - M + t)(J- M + ~)J1/2
C~~ -M - ~ i) (2J + 4)(2J + 3)(2J + 2)(2J + 1)
(J + ~, J,~)
J 2
J.) (_l)J-M+ 1/2[3(1 - M + t)(J - M + ~)(J + M + ~)J 1/2
c~~ 1 1.
-M -2 2 (2J + 4)(2J + 3)(21 + 2)(2J + 1)
I i) (_l)J-M-1 12[3(J - M - !Xl - M + t)(l + M + })J'2
czt -M-} } (21 + 3)(21 + 2X2J + 1)2J
(I + t, I,})
I }) J-M-112[ 1- M + t ll2 3
czt -M-t t (-1) (21 + 3)(21 + 2)(21 + 1)21 (J + 3M + 2)
I 2) 1 J-M[(l - M - 1)(1 - M)(l - M + 1)(1 - M + 2)J'2
(I ~ 2
-M-2 2 (- ) (2J + 5)(21 + 4)(21 + 3X2J + 2)(21 + I)
I 2( _1)J-M[(l + M + 2)(J - M + 2)(1 - M + I)(J - M)J'2
(I ~ 2 (I + 2, I, 2)
-M-I ~) (2J + 5)(2J + 4)(2J + 3X2J + 2X2J + 1)
I 2) (_I)J-M[6(1 + M + 2)(1 + M + 1)(1 - M + 2)(1 - M + 1)J'2

(I ~ 2
-M 0 (21 + 5)(2J + 4)(2J + 3)(2J + 2)(2J + 1)
I 2) 2( _l)'-M+ 1[(1 - M - I)(J - M)(J - M + I)(J + M + 2)J'2 <

C~l -M -2 2 (21 + 4)(21 + 3)(2J + 2)(21 + 1)2J N
(j
J 0
J-M+l [ ( J - M + 1)(J - M) ll2 :3
C~1 (-1) 2(J + 2M + 2) (2J + 4)(2J + 3)(2J + 2)(2J + 1)2J (J + I, J, 2) cr
-M-I ~) S·
~
J 2) IJ-M+1 [ (6J+M+l)(J-M+I) ll2 o·
(J ~1 ::l
-M 0 (- ) 2M (21 + 4)(2J + 3)(2J + 2)(2J + 1)2J
0
....,
1 J-M[6(J - M - 1)(J - M)(J + M + 1)(J + M + 2)J'2 tTl
J 2) 0-
M -M - 2 2 (- ) (2J + 3)(2J + 2)(2J + 1)(2J)(2J - 1)
C :3<1>
::l
I J-M [ 6(J + M + I)(J - M) J'Z ~
'"1
(J, J, 2) '<
-M-I ~) (-1) (l + 2M) (2J + 3)(2J + 2)(2J + 1)(2J)(2J - I)
~ :;:tl
0
J 2) J-M 2[3M Z - J(J + I)] ~
(~ -M 0 (-I) [(21 + 3)(2J + 2)(2J + 1)(2J)(2J _ 1)]112 '0
'"1
C/O
- ..I
-
Vl
V.3 Tensor Operators and the Wigner-Eckart Theorem
We have considered the angular-momentum operators in some detail in the

previous section. Now we consider a more general class of operators, called
tensor operators,6 which are defined by their relationship with the angular-
momentum operators. The simplest example of such an operator is a scalar
operator, which is defined to be any operator S satisfying
[1;, S] = 0. (3.1)
Another example is any set of three operators V; (i = 1, 2, 3) that satisfy
(i,j, k = 1,2,3). (3.2)
Such a set of operators is called a vector operator or regular tensor operator.
We note that the angular momentum itself is a vector operator.
Rather than using the" Cartesian" components Vi' it is more convenient
to use the" spherical" components Vo , V±1 defined by
(3.3)
Note that the H ± of (III.3.l) (or the J ± = hH ±) are not the spherical com-
ponents of a vector operator, but differ from them by the factor + 1/J2 or
-1/J2; therefore we use the notation J ± 1 for the spherical components
of the angular momentum operator, in contrast to J ± for the raising and
lowering operators. Equation (3.2) becomes
(K = -1,0, + 1), (3.4a)
[J o , VK ] = KhV±1'
(3.4b)
In general, we define a tensor operator of rank j to be (in spherical com-
ponents) a set of 2j + 1 operators 7 T~) (K = - j, - j + 1, ... , +j) that
satisfy
[Jo, T~)] = KhT~),
[J ±, T~)] = Jj(j + 1) - K(K ± l)hT~L. (3.5)

This can be written in a more compact form using the Clebsch-Gordan
coefficients:
[JI" T~)] = Jj(j + 1)(j K 1 pJj K + p>T~L (3.5')
6 We shall actually only consider what are called" irreducible tensor operators."
7 We emphasize the use of spherical components by the use of Greek letters such as K in
labeling the components. While such emphasis is not needed for the general tensor operator,
in the case that the tensor operator is a vector operator the use of Greek letters will enable us
to distinguish between the Cartesian components (Latin letters) and the spherical components
(Greek letters).
We note that scalar and vector operators are tensor operators of rank 0 and
1, respectively.
The matrix elements of tensor operators have an important property which
is expressed by the W igner-Eckart theorem:
Let T'!) be a tensor operator. The matrix element ofT~) between the angular
momentum eigenstates may be written as
<Im'l T'!)1j m) = (j m J KIIm') <III T(J)IIj), (3.6)
where (j m J KII m') is a Clebsch-Gordan coefficient and the symbol
<IIIT~)IIj) [defined by (3.6)J denotes a quantity that depends onj',j, J and
the nature of the tensor operator T'!) but does not depend on m, m', or K.
<III T(J)II j) is called the reduced matrix elementS of the tensor operator.
This theorem will not be proven here, since the proof is pure mathematics
and does not give any additional insight into the physics. If in addition to j
and m there are other quantum numbers, say YJ = (al' a2"'" aN), then the
reduced matrix element will in general also depend upon these quantum
numbers, i.e.,
<YJ'j'm'IT'!)IYJjm) = (jmJKIj'm')<YJ'j'IIT(J)IIYJj)· (3.6')
The additional observables YJop = (A 1, A 2 , ••• , AN) whose eigenvalues are

the quantum numbers YJ = (al, a z , ... ,aN) must have the property that
['lop, J;] = 0 or [YJ oP, SO(3)J,] = 0.1t should be emphasized that the Wigner-
Eckart theorem is both a theorem and a definition. It is a theorem in that
(3.6) says that the matrix element of T~) can be factored such that the
dependence on m, m', and K is entirely contained in the Clebsch-Gordan
coefficient, and it is a definition in that (3.6) defines the reduced matrix
elements.
The Wigner-Eckart theorem has become one of the most important tools
for the understanding of physics. Equation (3.6) is the Wigner-Eckart
theorem for the rotation group, which is connected with the algebra of
angular momenta J i as an algebra of observables. Many physical systems
have an (enveloping) algebra of a group as a subalgebra of the algebra of
observables, and they have observables that are tensor operators with
respect to this group. For those observables the Wigner-Eckart theorem,
a generalization of (3.6), holds.
The Wigner-Eckart theorem expresses the matrix elements of the tensor
operators, which are directly connected with the numbers measured in an
experiment, in terms of Clebsch-Gordan coefficients and reduced matrix
elements. The Clebsch-Gordan coefficients are known mathematical
quantities and are calculated from the properties of the group. The reduced
matrix elements are physical parameters whose values have to be obtained
8 In some textbooks, a constant factor or function of j appears explicitly in the Wigner-
Eckart theorem. In our notation these factors have been absorbed into the reduced matrix
element.
178 V Addition of Angular Momenta~ The Wigner-Eckart Theorem
from the experimental data. [If an observable has further additional proper-
ties-e.g., Equation (VI.3.12) of the Lenz vector in Chapter VI -then the
reduced matrix elements can be reduced to a still smaller number of param-
eters.] Thus the importance of the Wigner-Eckart theorem is that it allows
one to fit the large number of experimentally observable matrix elements in
terms of a much smaller number of more fundamental quantities, the reduced
matrix elements. The property of being a tensor operator is often all one
knows of an observable, and the Wigner-Eckart theorem then is the only
tool at hand.
A particular consequence of the Wigner-Eckart theorem (3.6) is
<j'm'l T~)1j m) = 0
if m' =1= K+ m or j' =1= J + j, J + j - 1, ... , IJ - j I (3.7)
which follows immediately from the properties of the Clebsch-Gordan
coefficients (2.33).
As an illustration of the Wigner-Eckart theorem, we note that for a scalar
operator S, Equation (3.6) becomes
(3.8)
This says that S cannot change the angular-momentum quantum number,
which is just as we would expect. For the angular-momentum operators
themselves, Equation (3.6) gives
<j'm'IJKU m) = <j m 1 Klj'm') <fllJllj), (3.9)
where the Clebsch-Gordan coefficients are listed in Table 2.2. By comparing

the Clebsch-Gordan coefficient in the table with the formula (III.3.8'),
(III.3.22), and (III.3.23), we see that
<IIIJIU) = (5j'jhJj(j + 1). (3.10)
If Ij m) is a basis in the space in which v" is a general vector operator, then
we can expand VK Ij m) with respect to this basis:
VKIj m) = I Ij'm') <j'm'!VKlj m)
and obtain, according to the Wigner- Eckart theorem (3.6),

v"ljm) = I lj'm')Gm 1 Klj'm') <j'llVllj). (3.11)
j'm'
If the c.s.c.o. contains, in addition to J2 and J 3, N other operators A l , ... , AN

with spectrum 1'/ = (a l , a2, ... , aN)' then the basis is II'/j m), and instead of
(3.11) one has, according to (3.6'),
~II'/j m) = I II'/j'm')G m 1 KIj'm') <1'/j'llVlll'/j) (3.11')

q'j'm'
We shall for the remainder of this section suppress the additional quantum
numbers 1'/, but it should be understood that whenever additional quantum
numbers fJ are needed, the reduced matrix elements will depend upon these
quantum numbers and summation over fJ' is implied.
According to (3.7), the only non vanishing terms in (3.11) are those for
whichf = j + 1,j,j - 1 and m' = m + K:
VKljm) = Ij - 1 m + K)(jm 1 KIj - 1 m + K)(j - 111V11j)
+ Ijm + K)<jm 1 Kljm + K)<jllVllj)
+ Ij + 1 m + K)(jm 1 Klj + 1 m + K)<j + 111V11j).
(3.12)
This is the most general form possible for the action of a vector operator.
According to (3.12), then, any vector operator can be completely specified
by three reduced matrix elements (which in general may depend upon
fJ, fJ')· Actually, as we shall show below, only two quantities are needed to
determine a vector operator.
By using the Clebsch-Gordan coefficients of Table 2.2, we can write (3.12)
explicitly. For the O-component,
Voljm) = Ij - 1 m)( - J(j ~(;)~ ~ m))(j - 111V1Ij)
+ Ij m)(J m )(jll VIU)

j(j + 1)
+ Ij + 1 m) (
(j - m + 1)(j + m +
(2j + 1)(j + 1)
1)) (j + 111V1Ij),
or
Volj m) = J / - m 2cjlj - 1 m) - majlj m)
- J(j + 1)2 - m2djlj + 1 m), (3.13)
where the Cj' aj' and dj are defined by Equation (3.13) itself:
(j - 111V11j)
Cj = - -J---r='j(=2j=+=1)=--'
(j11V11j)
a·= - - r = = (3.14)
J Jj(j + 1)'
d. = _ (j + 111V1Ij)
J J(2j + 1)(j + 1)
Likewise,
V+ 1 Ijm) = -Ij - 1 m + l)J(j - m - 1)(j - m)/2cj
+ Ijm + l)J(j + m + 1)(j - m)/2aj
- Ij + 1 m + l)J(j + m + 1)(j + m + 2)/2dj (3.15)
and
V_ 1Ijm) = -Ij - 1 m - I)J(j + m)(j + m - 1)/2cj
-Ijm - I)J(j - m + 1)(j + m)/2aj
- Ij + 1 m - I)J(j - m + 1)U - m + 2)/2 dj. (3.16)
Equations (3.13), (3.15), and (3.16) say that a vector operator is completely
determined by the three functions Cj' aj' and dj (which may in general
depend upon ri', 1]) of the discrete parameter j. In the mathematical note
below it will be shown that in fact only two such functions are needed, as
it is possible to choose the basis Ij m) in such a way that
(3.17)
i.e.,
J2j + 3<j + IIIVIIj) = J27+lGIIVIU + 1). (3.17')
[We define a new basis

Ih~) = w(j)ljm), (3.18)
where wU) is a complex number. The angular-momentum operators
have the same form (Section III.3) in the h~) basis as they do in
1
the Ij m) basis. In general, the Ih~) are not normalized unless

Iw(j)1 = 1. (3.19)
In this basis, Equation (3.13) becomes
Volh~) = J / - m 2cj (~(j) ) Ih~-1)

wJ-l
- majlh~)
-J('+1)2- 2d. w(j)_lh j + 1) (3.20)

J m J w(j + 1) m •
We define Aj = aj' and we wish to choose w(j) such that

w(j) w(j)
Cj w(j _ 1) = Cj and d j w(j + 1) = Dj = C j + l' (3.21)
This will be possible if

w(j) wU - 1)
Cj w(j _ 1) = w(j) d j _ 10 (3.22)
i.e., if
(3.23)
Suppose jo is the smallest value of j in the space, i.e.,

Go - 111V1Uo) = o. (3.24)
Y.3 Tensor Operators and the Wigner-Eckart Theorem 181
Then (3.23) will be satisfied if we choose
w(J') -_ W 2(')
Jo -djo- -
djo - dj -
+ 1 ... - - 1
(3.25)
Cjo+! Cjo + 2 Cj
so (3.20) becomes
Volh~) = J/ - m2Cjlh~-I) - mAjlh~)
-J(j + 1)2 - m2Cj+llh~+1).] (3.26)
Thus every vector operator can be written in the form (3.13), (3.15), (3.16)
with (3.17), where the Cj and aj are functions ofj.1t should be noted, however,
that in general we may have dj = Cj+ 1 for only one vector operator V" at a
time. If two different vector operators are involved in one problem and the
basis has been chosen so that for the reduced matrix element of one of them
(3.17) holds, then in this basis, the other or any additional vector operator
is expressed in terms of three independent reduced matrix elements.
v:
So far we have not required anything of except that it be a vector opera-
tor. As a consequence the reduced matrix elements Cj and aj are arbitrary
functions ofj.lfwe specify V" further we will obtain more specific information
about the Cj and aj. In the Appendix we will study three specific cases of
vector operators which together with the J" will generate the algebra of a
group, and we will see that in each of these cases the functions Cj and aj are
completely specified by a pair of two numbers. But if nothing else is known of
v: (from physics) except that it is a vector operator then the Cj and aj cannot
be further calculated. They can only be determined phenomenologically
from the experimental value of one particular matrix element of one com-
ponent of V"' e.g., c j can be obtained from (j - l,jlVoljj) and aj from
<jjl Voljj)· All other matrix elements (j mlVlj m) for m = -j, -j + 1, ... ,j
and K = 0, + 1, -1 can then be calculated from these two experimental
values using the Clebsch-Gordan coefficients.
Often the vector operators have a specified Hermiticity property. They
are either Hermitian, defined by
V6 = VO , VL = -V+l'
or they are skew-Hermitian, defined by
V6 = - VO , VL = + V+!.
It is easy to see (Appendix) that for Hermitian V" the functions Cj are purely
imaginary, Cj = -Cj' and the functions aj are real, ii j = aj. For skew-
Hermitian v: one has cj = Cj and ii j = - a j .
Appendix to Section V.3-Representations of the Enveloping

Algebras SO(3,1), SO(4), and E3
Algebras that are generated by the angular-momentum operators J i and a
particular vector operator are needed for various problems of quantum
physics. We have already seen an example of such an algebra in Section 111.4,
where <9'(E3) was the algebra generated by the angular momentum J i and
the dipole operator Qi' In Chapter VI we shall encounter the algebra <9'(SO( 4»
generated by the angular-momentum operator Li and the Lenz vector Ai'
<9'(SO(3,1» is the algebra of the Lorentz group and has various applications
in quantum physics.
In this appendix we give a derivation of the representations of these
three algebras. These algebras are the enveloping algebras of the groups
SO(3,1) [the pseudo-orthogonal group in (3 + 1) dimensions], SO(4) [the
orthogonal group in four dimensions], and E3 [the three-dimensional
Euclidean group], but we shall not discuss these group-theoretic connections
here.
The defining relations of these algebras are given by
[Hi' H j]= ifijkHb (3.27)
[Hi' F j] = ifijkFb (3.28)
[Fi' F j] = J..hfijkHk, (3.29 A2)
where J..,z = -1 for <9'(SO(3,1»,},,2 = + 1 for <9'(SO(4», and},,2 = 0 for <9'(E3)'
Equation (3.27) defines the algebra <9'(SO(3» of angular momentum. Equa-
tion (3.28) specifies that Fi is a vector operator with respect to <9'(SO(3»;
the F i , i = 1, 2, 3, are the Cartesian components. Equation (3.29) then
specifies that this vector operator together with Hk generates one of the above
algebras. If in addition to (3.27) and (3.28), (3.29) is also fulfilled, then
F is a very particular vector operator and its reduced matrix elements
have very particular values. We shall now determine these values. We shall
derive the representations of (3.27)-(3.29) for },,2 = -1 and obtain the other
two cases}" 2 = 1 and}" 2 = 0 by a change of the Hermiticity property and by
a limiting process, respectively.
To find all linear representations of the algebra (3.27)-(3.29) means to
find all linear operators in all linear spaces that fulfill these commutation
relations. We shall restrict ourselves here to the subclass of those representa-
tions whose representation space f!ll contains each of the representation
spaces f!lll (l = 0, or t, or etc of <9'(SO(3» at most once. 9 In each f!lll we have,
according to Section III.3, the basis
f~, m = -I, 1 + 1, ... , + I, (3.30)
and the space f!ll is then spanned by these basis vectors f~, where 1 runs
through a set of values that is to be determined. In other words,
(3.31)
If one of the f!lll, say for 1 = 10 , were to appear more than once, then f~
would not be a basis system of the space f!ll, and a new label (quantum
9 These include all the representations that are connected with unitary representations of
the groups SO(3, 1) and SO(4).
number) would be needed to distinguish orthogonal vectors with the same

values of 10 and m:f:::~.
The action of the linear operators H k on all f~ is already known from
Section 111.3, and we have now to determine the action of the Fk upon the
f~. We introduce the components [in analogy to (III.3.1)J
F± = F 1 ± iF 2 = +J2F ± I' (3.32)

F 0 and F ± 1 are the standard spherical components of the vector operator
F. The commutation relations of the F 0, F ± with H ± , H 0 = H 3 and with
each other are then
[H+,F-J = 2F o , [H~,F+J: -2FO}
[FK' HKJ = 0, K -0, +, (3.33)
[Ho,F±J=±F±, [H±,FoJ=+F±
[Fo,F±J = +H±, [F + , F _J = - 2H 0 . (3.34)
Equation (3.33) corresponds to (3.28) and expresses the vector-operator
property of F, and (3.34) corresponds to (3.29) with ,.1,2 = -1 and specifies
that this vector operator is a generator of @,,(SO(3,1)). If we now use the
vector-operator property and the Wigner-Eckart theorem, then we obtain
for F 0 and F ± the expressions (3.13), (3.15), and (3.16). If we also choose the
phase factor of the basis vectors appropriately, as discussed in the mathe-
matical note just before this appendix, then (3.17) can be used, and we
obtain for the action of F 0 and F ± on the basis vectors the expressions
Fof~ = JI 2 - m2cd~-1 - mad~ - J(l + 1)2 - m2cl+lf~+I, (3.35 0 )
F + f~ = J(l - m)(l- m - l)cd~-+\ - J(l- m)(l + m + l)aJ~+l
+ J(I + m + 1)(1 + m + 2)CI+ d~++\, (3.35+)
F _ f~ = - J(I + m)(1 + m - l)cd~--\ - J(I + m)(I- m + l)ad~-1
- J(I - m + 1)(1- m + 2)CI+1f~+-\. (3.35_)
These have been obtained using only (3.33); and every vector operator, in a
space in which PAl appears at most once, has this form.
The vectors j~ have, according to the results of Section 111.3, the property
H3f~ = mf~, H2f~ = l(l + l)f~,
(3.36)
H±f~ = J(I + m)(l ± m + l)f~±I'
The f~ have in addition been chosen so that (3.17) is fulfilled.
So far we have only exploited (3.27) and (3.28); we shall now make use of
(3.29) or (3.34) to determine the unknown coefficients al and Cl [related to
the reduced matrix elements by (3.14)]. It is sufficient to use just one of the
three relations (3.34) because the other two are consequences of it and the re-
lations (3.33). We choose
(3.34')
Applying both sides of (3.34') to the vector f~, using (3.35), and comparing
coefficients of f~+ 1,J~! L and f~-+\ leads to the following equations:
[all + 1) - (I - 1)al_l]cl = 0, (3.37a)
[al+ 1(1 + 2) - laaCl+ 1 = 0, (3.37b)
(21 - 1)cr - (21 + 3)cr+ 1 - ar = 1. (3.37c)
One starts with an arbitrary basis vector f~ which one assumes is contained
in the particular representation space f3l that one wants to construct. Apply-
ing the operators H K to f~ a sufficient number of times, one obtains, as
described in Section 111.3, the whole space ~I. Applying the vector operators
FK to f~ a sufficient number of times, one reaches all the other ~j c ~ where
j differs from I by an integer, because F K changes I by 0, + 1, -1. As j > 0,
there must always exist a smallest value of j in any f3l. This smallest j we call
ko ; it can be any ofthe allowed values for j and will characterize the space f3l.
If ko is the smallest value of j, then according to (3.35)
Cko = 0.
If ko is integer, then f3l contains only integer I ;::: ko, and if ko is half-integer,
then f3l contains only half-integer I ;::: ko :
f3l= (3.38)
I=ko,ko+ 1, ...
For alii for which CI # °one obtains from (3.37a)

all + 1) - (I - 1)al-l = 0, (3.39a)
and for alii for which CI+ 1 # ° one obtains from (3.37b)
al+ 1(I + 2) - la l = 0. (3.39b)

Defining
PI = l(l + 1)aj, (3.39c)
these two conditions lead to
PI - PI-l = 0, (3.39d)
Pl+l - PI = 0,
which means that PI is independent of l. We call this arbitrary complex
constant PI = ik o c, where C is an arbitrary complex number, and we shall
show below that the factoring out of ko is always possible. al can then be
written
ikoc
a[ = l(l + 1)
for any I;::: ko, I ;::: t. (3.40)
af =
°
For the case that both CI = and CI+ 1 = 0, it follows from (3.37c) that
-1, and the matrix elements (3.35) of F K are completely determined.
With (3.40) this means that ko = I, C = I + 1 and that f3l = f3l1.
°
It remains to show that the factoring out of ko is always possible, i.e.,
that al = for ko = 0. But for ko = 0, (3.39a) holds for 1 = 1 and leads to
a 1 = 0. For 1 = 2 this in turn leads to a2 . 3 - al = a2 . 3 = 0, and proceed-
ing this way we obtain a l = 0 for 1 = 1, 2, 3, .... For I = 0 the factor in
front of ao in (3.35) is zero so that the appearance of ao is meaningless. Thus
(3.40) holds generally.
To determine CI we use (3.37c). Defining
(JI = (21 - 1)(21 + l)c;, (3.41)

we can write (3.37c) as
(JI - (Jl+l - (21 + l)a; = 21 + 1,

or using (3.40)
(JI - (JI+ 1 = (21 + 1) - 2 2

koC (1r - +1)
(l 1)2 . (3.42)
We calculate now for any value k ;::: ko, k ;::: 1
= k2 - ko2 - ko22(1
c k~ - k1)
2
(k 2 - k~)(P - c2 )
k2
As ko is the smallest value of I, Cko = 0, and consequently (Jko = 0, so that

one obtains
(3.43)
and by (3.41),
w- k~)W - c2 )
for any 1;::: ko, 1 ;::: 1. (3.44)
412 - 1
Therewith, we have found all possibilities of how the operators H" and F",
that fulfill (3.33) and (3.34), can act upon the vectors f~. These possibilities
are characterized by the pair of numbers (ko, c), where ko is an integer or
half-integer and c is a complex number. For each choice of these two numbers
the action of the operators on the f~ is given by (3.35) and (3.36), with al
and CI given by (3.40) and (3.44). In other words, for each pair (ko, c) we have
a representation of the algebra 0"(SO(3,l)) generated by the H" and F" in
the representation space f!Il of (3.38). This space and the operators acting in it
are characterized by the pair of numbers (ko, c):f!Il(ko, c), H~ko.C), and F~ko.c).
So far, we have not introduced into fJ£(ko, c) a scalar product and cannot
speak of Hermiticity of the operators. Therefore, the symbol LI=ko.ko+ I .... EB
defines the space of linear combinations
f = po + pO+1 + ... + fl + ... + fN,
with fl E fJ£1 and N in general larger than any given number. For all of the
pairs (k o , c), the H~ko, c) and F~ko, c) are all of the linear representations of
g(SO(3,1» that contain a given value of angular momentum 1at most once.
These representation spaces are in general infinite-dimensional; only for
certain values of c are they finite-dimensional.
The finite-dimensional cases are obtained from (3.35) and (3.44). Looking
at (3.35), we see that successive application of F K will lead to ever higher
values of I unless for some value 1 = kl one has Ckl + 1 = 0; kl is then the
highest value that I can take in fJ£. According to (3.44), this means that the
value of c must be such that c 2 = (k l + 1)2. Thus for
ko = 0, t 1, ... , and c = ±(k l + 1), where kl = ko, ko + 1, ... ,
(3.45)
the representation space is finite-dimensional and is given by
(3.46)
I=ko,ko + I, ...
Though fJ£(ko, (k l + 1» and fJ£(ko, - (k o + 1» are the same direct sum of
irreducible representation spaces of g(SO(3», the F K act differently in these
two spaces, as can be seen from (3.44) and (3.35). They are inequivalent
representation spaces of g(SO(3,1», but they have the same reduction with
respect to g(SO(3». From (3.40) and (3.44) it follows that for the finite-
dimensional cases (3.45) the al and CI fulfill the conditions
(3.47b)
We will now introduce a scalar product into the spaces fJ£(ko, c), of which
we must require
(f~,J~,) = bll'b mm, (3.48)
(so that H2 is a Hermitian operator in fJ£(ko, c». Then from (3.35 0 ) one
obtains
(Fof~-l,f~) = -J1 2 - m 2 ctCf~,J~),
(f~-I,Fof~)= J[2 - m2cI(f~-I,J~-I),
and
(Fof~,J~) = -mal(f~,J~),
(f~, Fof~) = -matCf~,J~)·
With (3.48), this means that
°
(F f~- 1,J~)CI = - °
(f~-1, F f~)cl>
and
From (3.50) and (3.47b) it follows that for the finite-dimensional cases (3.45)
°
the operator F must be skew-Hermitian: F6 = - F o. By similar arguments,
or by (3.34), it can also be shown that the other components of Fare skew-
Hermitian. Thus F is a skew-Hermitian vector operator,
(3.47a)
if the representation is finite-dimensional.
The other very important case is that F is a Hermitian vector operator:
(3.51a)
i.e., that we have a Hermitian representation of g(SO(3, 1)) [which is con-
nected with a unitary representation of the group SO(3, 1)]. In this case,
it follows from (3.51a) and (3.50) that
(3.51b)
This means according to (3.40) that ic must be real unless ko = 0, and
according to (3.44) that W - k~)W - c2)j(412 - 1) must be positive. The
latter is possible only when c 2 is real, i.e., when c is real or purely imaginary.
Thus for any integer of half-integer value of ko and any purely imaginary
value of c, (3.51) can be fulfilled. If c is real, then ko must be equal to zero in
order that al = al' Then the expression under the square root in (3.44)
becomes 12([2 - c 2)/(4F - 1), which will be positive for all 1 = 1, 2, ...
only if c 2 < 1.
Thus we have seen that (3.51) can be fulfilled in two cases:
For ko = 0, t 1, ... and ic = real, -
a.
b. °
For ko = 0, ~ c < 1.
C1J < ic < + C1J
In neither of these two cases can Cl+ 1 ever become zero. Consequently,
according to (3.35), one obtains from any given f ~ also an f~+ 1. One can
apply F I( an arbitrary number of times arriving at ever higher values for I.
The representation space flll = &f(ko, c) with (ko, c) fulfilling case a or b is
infinite-dimensional:
00
L
l=ko,ko+ 1, ...
(3.52)
It is the (algebraic)10 direct sum of the irreducible representation spaces

&fl(l = ko, ko + 1, ... ) of the algebra of angular momentum, g(SO(3)).
(k o = 0, c = 1) is the trivial case in which all HI( and FI( are represented by
the zero operator.
10 It can be completed with respect to various topologies, e.g., with respect to the Hilbert-
space topology defined by the scalar product, or with respect to the topology defined by the
countable number of scalar products (</>, 1/1)_ = (</>, (H 2 + F2 + 1)-1/1) leading to the nuclear
space <1>. Cf. B6hm (1978).
Summarizing our results, we have found that the irreducible-representa-

tion II spaces of 6"(SO(3, 1» [the algebra defined by (3.27), (3.28), (3.29 A2 = - I)J
for which Hi and F; are Hermitian are characterized by two numbers
(ko, c) that can take any of the values in case a or b. The reduction of these
representation spaces ~(ko, c) with respect to the algebra of 6"(SO(3» is
given by (3.52), and the operators H" and F" are given by (3.36) and (3.35)
with (3.40) and (3.44). The irreducible representation spaces of 6"(SO(3, 1»
with skew-Hermitian Fi are characterized by the two numbers (ko, c) =
(ko, ±(k l + 1» which can take any of the values (3.45). The reduction of these
representation spaces with respect to 6"(SO(3» is given by (3.46), and H" and
F" are again given by (3.36), (3.35), (3.40), (3.44).
It is now easy to obtain the representations of 6"(SO(4» and 6"(E3)' To
obtain the representations of 6"(SO(4» we define
A" = -iF" (K = 0, +, -). (3.53)
If F j fulfills the commutation relation (3.29 A2 = -1) then Ai fulfills the com-
mutation relation (3.29 A2= + 1); i.e., it fulfills together with Hi the commuta-
tion relations of 6"(SO(4»:
If F j are skew-Hermitian (FJ = -F), then Aj are Hermitian (AJ = A).

Therefore we conclude that the Hermitian irreducible representations of
6"(SO(4» are characterized by two numbers (ko, ±(k 1 + 1» that can take
the values (3.45). The irreducible-representation spaces are all finite-dimen-
sional and given by (3.46). The operators A" are again given by (3.35) with
(3.40) and (3.44), where F" is given by (3.53).
The algebra 6"(SO(3, 1» has two independent invariant operators, i.e.,
two independent operators that commute with all generators Hi and with
F i • They are
C 2 =F·H. (3.54)
As they commute with all H" and F" their eigenvalues cannot change by
applying H", F" to the corresponding eigenvector. Therefore their eigen-
values must be the same in the whole irreducible representation space. One
can calculate these eigenvalues applying C 1 and C 2 to a suitably chosen
f~ using (3.36), (3.35) with (3.40) and (3.44). The result is
Clf~(ko,c) = (-k6 - c 2 + 1)f~(ko,c),

C2f~(ko, c) = ikocf~(ko, c).
The two invariant operators for 6"(SO(4» are

C~O(4) = A . H, (3.56)
11 The irreducible representations are also called ladder representations because they are
obtained by climbing from one value of I to the next.
with the eigenvalues

C~O(4)f~(ko, ±(k 1 + 1» = + c2 - l)f~(ko, ±(k 1 + 1»
(k6
= (k6 + ki + 2kdf~(ko, ±(k 1 + 1», (3.57 1 )
C~O(4)f~(ko, ±(k 1 + 1» = kocf~(ko, ±(k 1 + 1»
= ±ko(k1 + l)f~(ko, ± (k1 + 1». (3.57 2 )
Thus, the eigenvalues of the invariant operators specify the irreducible

representation completely; it is, however, more convenient to characterize
the irreducible representations by the values of (ko, c) or by the values of
(ko, k1' ±), where ± refers to the two inequivalent representations with
the same value of ko and k 1.
We finally mention the special case where ko = 0 or C 2 = O. This is the
case one encounters most frequently in the application to simple quantum-
mechanical systems. We shall meet this case for 0"(SO(4» again in Chapter
VI in the derivation of the hydrogen spectrum. It is the case that does not
require parity doubling, as we shall see below in Section V.4 [in particular
Equation (4.25)]. We have encountered it already in Section 111.4; it leads
to a spin spectrum given by (111.4.2) and depicted by Figure 111.4.1. For
ko = 0 it follows from (3.52) for Hermitian representations of 0"(SO(3, 1»
that
~(O, c) = L EEl ~l, (3.58)
1=0.1 • ...
and from (3.46) for Hermitian representations of 0"(SO(4» that

n-1
~(O, k1 +1= n) = L EEl ~l (n = 1, 2, 3, ... ). (3.59)
1=0.1 • ...
From (3.40) follows that for this case

al = 0 for alII (or </IIFII/) = 0) (3.60)
and the F K and AK change the value of I by ± 1. Therefore, the representa-
tions (0, k1 + 1) and (0, -k1 - 1) are equivalent.
We shall now derive the irreducible representations of 0"(E3) from the
irreducible representations (k o , c) of 0"(SO(3, 1» by a limiting process,
called Inonii-Wigner contraction. We define the operators
(3.61)
If Hi and Fi fulfill the commutation relations (3.27), (3.28), (3.29;'2= -1) of
0"(SO(3, 1», then Pi and J i fulfill the commutation relations
= ifiklJl,
[J i , J k] (3.62)
[J;, P k] = ifiklPl, (3.63)
[Pi' P k] = -)..2fiklJi> (3.64)
and Pi depends upon A. The invariant operators are now

A2C 1 = PiP i - A2 JJi' (3.65)
AC 2 = PJi· (3.66)
In the limit A2 -+ 0 the commutation relation (3.64) goes into the commuta-
tion relation
(3.67)
Equations (3.62), (3.63), and (3.67) are the commutation relations that define
the algebra ,$(E3). Thus in the limit A2 -+ 0 the commutation relations of
,$(SO(3, 1» go into those of ,$(E3). However, if we simply take the limit
A -+ 0, then we see from (3.61) that Pi -+ O. In order to obtain the algebra of
operators of ,$(E3) from the algebra of operators ,$(SO(3, 1» in rJf(ko, e)
by the limit A -+ 0, we must simultaneously increase the F K so that P K does not
go to zero. This can be done, as seen from (3.35), (3.40) and (3.44), by in-
creasing the value of Iie I -+ 00 when A -+ O. The contraction in the represen-
tation is therefore performed in the following way:
A -+ 0, liel -+ 00 such that id -+ f (3.68)
where f is a finite real value. From (3.35), (3.40), and (3.44) we see that, e.g.,
P 3 is then given by
P3f~ = lim AFof~
where
[2 - k6
Cz = lim Ac z = lim _Ii 4[2 - 1 (3.70)
iCA-£ iCA-i
and
- 1~ }..icko
az = 1·1m /LUz = 1·1m (3.71)
id~, id~, [(l + 1)
The invariant operators of ,$(E3) are PiP i and PJi. From (3.65) we see that
in this limiting process
PiP i = lim A2C 1 = lim (-k6}..2 - }..2C 2 + ,Fl) = f2, (3.72)
iCA-! iCA-£
where we have used (3.55 1 ) for the irreducible representation (ko, c). In the
same way it follows from (3.66), using (3.55 2 ), that
PJi = lim AC 2 = lim Aicko = fk o . (3.73)

iCA-f iCA-i
The representation of ,$(E3) obtained in this contraction process (3.68) is

therefore characterized by the two parameters (ko, f) with - 00 < f < + 00.
The $'(SO(3» subalgebra and the reduction with respect to this sub-
algebra are not affected by the contraction process (3.68). Therefore the
irreducible representation space ~(ko, E) of $'(E3) obtained from the rep-
resentation space ~(ko, c) is again given by [cf. (3.52)]
~(ko, E) = (3.74)
I=ko.ko+ 1 ....
Summarizing our results, we have found that the Hermitian irreducible-
representation spaces of $'(E3)' defined by (3.62), (3.63), (3.67), are charac-
terized by two numbers (k o, E) which take the values ko = 0, t, 1, ... , - 00 <
E < + 00. The reduction of this space with respect to $'(SO(3» is given by
(3.74), and the operators J = H K and PK are given by (3.36) and (3.35)
I(
with FI( replaced by PI( and al> CI replaced by iii' and CI given in (3.70) and
(3.71).
The basis system f~ of the representation space ~(ko, E) is the one in
which the following complete system of commuting operators is diagonal:
(3.75)
As the Pi commute [Equation (3.67)], one usually chooses a basis system
of the representation space in which the Pi are diagonal. As a complete
system of commuting operators one chooses therefore
(3.76)
with the corresponding basis vectors Ipi> ko), which have the property
<k~p;lpiko) = bkok~(P(P - p'), (3.77)
PdPikO> = PdPikO>' PJdPikO) = koElpikO)' (3.78)

E2 = PiPi'
Without a derivation,12 we mention that the transformation matrices
from the basis f~(ko, E) to the basis IPikO>'
IPikO> = If~(ko,E)<Ekolmlpiko), (3.79)
I.m
are given by
(21+11 I
(3.80)
<Ekolmlpiko> = './~~Dmko(<p,e, - <p).
Here (<p, e) are the spherical coordinates of p; P1 = lEI sin e cos <p, P2 =
IE I sin e sin <p, P3 = IE I cos e; and D~ko(<p, e,
- <p) is the rotation matrix. 13
In this appendix we have derived a huge class of irreducible representa-
tions of $'(SO(4», $'(SO(3, 1», and $'(E3)' many of which are used for ap-
plications in this book or for applications to other problems in quantum
12 A derivation is given in Appendix I of A. Bohm and R. B. Teese, Spectrum generating

group of the symmetric top molecule. J. Math. Phys. 17, 94 (1976).
13 The rotation matrices can be found in, e.g., L. C. Biedenharn and J. D. Louck [1979],
Chapter 3, or A. R. Edmonds [1957].
physics. We have used here the same method that was used in Section 111.3
for the derivation of the representation of 6"(SO(3)). Though group theory
has not been introduced, we want to mention that these are all the unitary
representations of the groups SO(4), SO(3, 1) (the homogeneous Lorentz
group), and E3 (the Euclidean group in three dimensions).
V.4 Parity
Parity is a very important observable in quantum physics, as all quantum-

mechanical systems are usually in states that are eigenstates of parity. The
parity operation P, which is also called space inversion, is the operation of
taking the mirror image. If a physical object has the position Xi and momen-
tum Pi' then its mirror image has the position xf = -Xi and momentum
pf = -Pi' For a quantum physical system this transformation P is, ac-
cording to the basic assumption I, represented by a linear operator in the
space of physical states. We call this operator the parity operator V p.
Some properties of V p follow directly ftom the physical interpretation
of the parity P; others are convention. As we shall justify below, V p is re-
quired to fulfill the following defining relations:
VpQiV;1 = -Qi (i = 1,2,3), (4.1)
UpP i U;1 = -Pi (i = 1,2,3), (4.2)
V pJ i V;1 = +J i (i = 1,2,3), (4.3)
where Qi' Pi, and J i are the position, momentum, and angular-momentum
operators, respectively. A further requirement on Up is
(4.4)
and
V;' = f. (4.5)
A consequence of (4.4) and (4.5) is
ut = ViI, (4.6)
Linear operators that fulfill (4.4) are called unitary operators ;14 thus Up is
unitary and according to (4.6) also Hermitian.
Equation (4.3) follows from (4.1) and (4.2) if J i = Li = tijkQjP k ; for other
angular momenta it is a postulate.
From (4.1), (4.2), and (4.3) we obtain the relation of V p with any ob-
servable that is a function of momentum, position, and angular momentum.
When new independent observables are introduced, their relations with Up
will have to be postulated.
14 See also Section XV.3 for the definition of a unitary operator.

V.4 Parity 193
The parity provides a classification of tensor operators. A tensor operator

n for which
k)
Upnk)Ui 1 = (-llT~k) (4.7)
is called a (proper) tensor operator; if

UpT~)Uil = _(_1)kT~k), (4.8)
then T~k) is called a pseudo tensor operator. Thus the operators Qi and Pi are,
according to (4.1) and (4.2), proper vector operators, whereas the operator
J i is a pseudo vector (also called an axial-vector) operator, according to (4.3).
All energy operators that we have met so far fulfill the condition
(4.9)
i.e., they are proper scalar operators. Equation (4.9) is, however, not uni-
versally fulfilled; the part of the energy operator that causes the weak decay
is known not to fulfill (4.9) (parity nonconservation).
In the following we shall give some justification of the above postulates:
Let ljJ be a vector in the space of states Yf, and denote by ljJR the vector we
obtain from ljJ by applying Up:
ljJR=UpljJ. (4.10)
Let us consider the pure physical state of which ljJ is the representative
vector, i.e., the state A"" which is the projector on the one-dimensional
space spanned by ljJ. Let us denote by A~ the projector on the space spanned
by ljJR, and let us make the "physically reasonable" assumption that this
is again a pure state (i.e., ljJR spans a one-dimensional space); ljJR is one
of the vectors that represent the state obtained from A", by a space reflection,
because Up is an operator that represents this space reflection. Thus A~ is
the space-reflected state of A"" and these two states are connected by
(4.11 )
As a generalization of (4.11) we obtain [cf. (1I.4.31)J that for any arbitrary
state W,
wR = UpWU~ (4.12)
will describe the state that is obtained by a space reflection of the state W.
When W is normalized (Tr W = 1), the space-reflected state should also
be normalized (Tr W R = 1), or more generally,
Tr W R = Tr W, (4.13)
i.e., space reflection should not change the probability. Thus
Tr W= Tr(UpWU~) = Tr(WU~Up),
which is fulfilled if one requires (4.4).
If one takes the mirror image of an object two times, then one gets the
original object back. Therefore
(4.14)
This is fulfilled if one requires (4.5), which is sufficient to ensure (4.14) but
not necessary for (4.14); i.e., (4.5) is also partially a convention.
Let Xi be the expectation value of the position operator of a quantum-
mechanical system in the state W,
Xi = Tr(Qi W);
and let xf be the expectation value in the state W R of the quantum-mechan-
ical system obtained from the state Wby space reflection, W R = Up WU; :
xf = Tr(Qi WR).
Then, because of the physical meaning of space inversion,
-R - (4.15)
Xi = -Xi;
i.e., space reflection changes the value of Xi into the value -Xi' Thus we
should have
Tr(Qi W) = - Tr(Qi WR) = - Tr(Qi Up WU~)
= -Tr(U~QiUpW), (4.16)
This is fulfilled if
or, using (4.4), if
which is the requirement (4.1). The other defining relations (4.2) and (4.3)
can be justified by similar arguments.
We have therefore justified the defining relations (4.1 )-(4.5) of the operator
that represents the observable parity by showing that these relations lead
to consequences that we expect of the physical operation of space reflection.
Thus Up defined by (4.1)-(4.5) can represent the observable parity. We have
not, however, shown that this Up is the only operator that can represent
parity. (A few remarks concerning this question are given in Section XIX.2,
in particular in the Appendix of that section.)
We shall now study the properties of the parity operator Up in angular-
momentum eigenstates-for example, as they occur in the rotator space
(111.4.2):
(4.17)
As the system of commuting observables
J 2 ,13 (4.18)
commutes with Up, it follows that
(4.19)
V.4 Parity 195
is also a system of commuting observables. We have to distinguish two cases:

(a) that (4.18) is already a C.S.c.o.-as is the case for the rotator-and (b)
that (4.19) is the c.s.c.o. This latter case is called parity doubling.
Case (a): If (4.18) is the c.s.c.o., then Ijj3) is a basis system. Because of
(4.3) it follows that
(4.20)
[n =f. 0 because of (4.4)], i.e., Uh) are eigenvectors of Up. The eigenvalue
n = n(j,j3), which we call the parity of the state Uh), could in general
depend uponj,h. We will now show that
(4.21)
where 11, called the intrinsic parity, is independent of j and h, and where
1111 = 1.
From (III.3.22) and (lII.3.23) it follows that 15
UpJ ± Uh) = upjj(j + 1) - ± 1)Uj3 ± 1)

j3(h
= n(j,j3 ± 1)jj(j + 1) - j3(j3 ± 1)ljj3 ± 1).
Because of (4.3) the left-hand side of this must be equal to
J ± U plj j3) = n(j,j3)J ± Uh)

= n(j,j3)jj(j + 1) - hCh ± 1)Uj3 ± 1).
Comparing the last two equations, we conclude that
(4.22)
i.e., that n is independent of j3' To find the dependence of n upon j, we have

to use (4.1) and the vector-operator property of Qi' Writing
and using (3.12), we have
QKljj3) = U - 1h + K)<jhl KU - Ij3 + K)G - 111QIID

+ Uh + K)<jh1 Kljj3 + K)GIIQIU)
+ U + 1h + K)Gj3 1 Klj + 1h + K)G + 11IQIU), (4.23)
15 In this and succeeding sections we shall use units such that h := 1. We shall on occasion
restore the h's to an equation, in which case we are using the usual (Gaussian) c.g.s. units, cf.
the explanation on p. III and on p. 209.
where </ II Q II j) are the reduced matrix elements of Q. If we apply Up to both

sides of (4.23), then from (4.1) it follows that
QK UPUj3) = n(j)QKUh) = -UpQKUh)
= -lj-1h+K)n(j-1)<jj3 1K lj-1j3+ K)(j-11IQIU)
- Uh + K)nU)(jh lKJjh + K)<jIIQIIj)
- Ij + 1h + K)n(j + 1)<jj31 Kij + 1h + K)(j + 11IQIU).
(4.24)
If we insert (4.23) into the second term of this equation and compare it with
the last term we conclude
(jIIQllj) = 0, (4.25)
n(j - 1) = - n(j) if (j - 111 Q II j) # 0, (4.26)
n(j + 1) = -n(j) if (j + 111QIIj) # O. (4.27)
According to (3.17'),16
J2j + 3 (j + 111QIIj) = J2j + l(jIIQIIj + I), (4.28)
so if <jlIQIIj + I) = 0 for every j, then this and (4.25) would mean that QK
is the zero operator. Thus for some value(s) of j, <jIIQIU + I) # 0; for all
such values we have by (4.26), (4.27)
n(j) = -n(j + 1). (4.29)
From this it follows that
n(j) = ( -l)j~ (~ = const). (4.21)
To determine what ~ can be we use (4.5):
UpUPUj3) = n(j)n(j)Uj3) = IUh), (4.30a)
and (4.4):
Thus
n(j)n(j) = 1, n(j)n(j) = 1, (4.31)
and consequently n(j) = n(j), so n(j) is real and n(j)2 = 1, i.e., n(j) is + 1 or
is -1. Thus
~ = +1 or ~ = -1. (4.32)
16 If the basis has not been chosen such that (3.17) is fulfilled for the vector operator Q but
if the Qi are Hermitian then one can prove that
-J2j + 3(j + lIIQIU) = J2j + 1(jIlQIU + I).

For skew-Hermitian Qi' i.e., Qt = - Q.. one can prove a similar relation with (-1) on the
left-hand side replaced by (+ 1).
V.4 Parity 197
We can now give a complete justification for the selection rule (I1IA.I0):
/1j = ± 1. (4.33)
It is a consequence of the fact that QK is a proper vector operator.
From the property
[J k> Ql] = iE klrn Qrn (k, I, m = 1, 2, 3) (4.34)
it follows that /1j = ± 1,0 [see Equation (3.7)]. From (4.25), i.e., the property
(4.1) of the Qi together with the assumption that there is no parity doubling,
it follows that /1j "# 0, which establishes (I1I.4.10).
Case (b): If (4.19) is the c.s.c.o., then the parity n is an additional label
for the basis vectors:
(4.35)
The question now is what is the spectrum of Up, i.e., what are the possible
values of n. From (4.4) and (4.5) it follows in the same way as above that (4.31)
must hold. Thus
n= +1 or n = - 1. (4.36)
Instead of (4.23) we now have
QKlij3 n ) = Ii - Ij3 + K, - n)Gh 1 Kli - Ij3 + K)(j - 1, -nIIQIUn)
+ Ijh + K, - n)(jh 1 Kljj3 + K)G, - nllQllin)
+ Ij + 1 j3 + K, -n)(jj3 1 KU + Ij3 + K)(j + 1, -nIIQlljn),
(4.37)
where in addition to the vector-operator property (4.34) of QK we have
taken (4.1) into account, i.e., that QK changes the eigenvalue of Up, and also
(4.36). Thus unlike case (a), (4.1) with (4.36) does not lead to any conditions
on the reduced matrix elements (j' n'llQllj n) and parities. Consequently
for every choice of j, j 3 there are two vectors
Ii h n = + 1) and [j h n = - 1). (4.38)
That is the reason case (b) is called the parity-doubling case.
It is customary to obtain an analogy with case (a) by labeling the states
not with n but rather with the intrinsic parity 1'/, which is defined by
(1'/ = ± 1). (4.39)
The parity n(j) in a state with angular momentum j and intrinsic parity 1'/
is then given by
n(j) = (-I)il'/. (4.21)
The selection rule is now seen from (4.37) to be
j~j+l,j,j-l, (4040)
n ~ -n. (4041)
Equation (4.41), which is a consequence of (4.1), was also fulfilled in case (a),
only there it was already implicit in the selection rule (4.33) as a consequence
of (4.21).
If parity is an independent observable (case (b)) then, according to (4.38),
for every value of j we have two angular momentum spaces
(4.42)
As QK transforms from a given value ofl' to (j + 1) - ",r '" and (j - 1) - ", one
obtains all angular-momentum spaces unless there is ajo such that
(4.43a)
or a j 1 such that
(4.43b)
That is,jo is the smallest value ofj that can be reached by repeated applica-
tion of QK' and j 1 is the largest. The space of all states may then be written
i1 i1
fYt = L EB fYt~= + 1 EB L EB fYt~= -1 = fYtn= + 1 EB fYtn= -1' (4.44)
j=jo j=jo
Are there physical systems in nature whose space of physical states is given
by (4.44)?
When we considered the rotating diatomic molecule, we used as its
classical picture the dumbbell, i.e., we assumed that the moment of inertia
about the line joining the two atoms was zero. However, because there are
a number of electrons revolving about the two nuclei, a better classical
model in many cases is a dumbbell with a flywheel on its axis. Thus in many
cases the diatomic molecule is not a simple quantum-mechanical rotator
but a quantum-mechanical symmetric top; the total angular momentum
j is no longer perpendicular to the direction of the figure axis a = x/I xl, but
instead has a constant component in the direction a, i.e.,
j. a = const, (4.45)
due to the revolution of the electrons.
~.- .... -.
\
\
\
\
\
\
\
\
\
\
\
\
\
\
: ~ ____ .....---"'L-~----.-J
a
Figure 4.1 Vector diagram for the symmetric top. The curved arrow indicates the
rotation of the whole diagram about j. The dashed part of the figure gives the vector
diagram when the sense of the direction of a is reversed.
V.4 Parity 199
To obtain the energy operator and the commutation relations of the

quantum mechanical symmetric top we use again the correspondence
between classical and quantum physics: We start from the classical non-
relativistic symmetric top17 with moment of inertia IA about the symmetry
axis given by the direction a and moment of inertia IB about any axis perpen-
dicular to a. As a rotates with the angular frequency co one has
a = co x a. From this one obtains-by taking the cross product with a
and using a 2 = 1-that
co = (co a)a 0 +a x a. (4.46)
Thus co has a component in the direction of a and one perpendicular to a.

The angular momentum j = 0 co, where 0 is the tensor of inertia, can be
0
written
j = IA(co a)a
0 + IBa x a.
Inserting again (4.46) this can be written
Using
x
aoj=aoO°co=IAaoco and a=1iI
one obtains from the equation above:
(4.47)
The energy of the top with angular frequency co and angular momentum j is
E= -!to 0j = -!to 0 0 0co.

Inserting (4.47) into this equation one obtains the energy of the classical
symmetric top in terms of j and x:
E = - 1 (02
J - IA - IB 21 (XOJ0)2) . (4.48)
2IB IA X
For the quantum-mechanical system, the numbers ji and Xi are replaced

by the operators J i and Qi. The energy operator of the quantum-mechanical
symmetrical top is therefore given by
(4.49)18
17 Corben (1968).
18 I/Q 2 is the operator defined as the inverse of Q2: I/Q 2 = (Q2) -1; see [M] in Section VI.3.
and the Qi and J i fulfill the commutation relations:
The solutions of these commutation relations, which are identical

with (3.62), (3.63), and (3.67), have been derived in the Appendix to
Section V.3. It has been shown that for every real number x and every
integer or half-integer jo there exists an irreducible representation
space which has the reduction with respect to angular momentum
given by (4.44) withjl = 00. The eigenvalues of the Casimir operators
Q2 and Q . J are given by
Q2 = QiQi = x 2 ,
(4.51)
Q.J = QJi =jox.
The irreducible representation spaces (4.44), in which the J i , Qj' and
Up fulfilling (4.50) and (4.1 )-(4.3) act as operators, are therefore
candidates for spaces of physical states of the symmetrical top. In
these spaces x 2 andjox are constants; they would therefore describe
symmetrical tops for which (4.45) is fulfilled, like the rotating di-
atomic molecule with electrons revolving about the two nuclei (as
long as the state of the electron does not change, which would change
the value ofjo)' For the general symmetric top (4.45) is not fulfilled
and its space of physical states is therefore a direct sum of the irre-
ducible representation spaces (4.44) with different values oUo. We
will not consider the general symmetric top molecules here. 19
The matrix elements of Qi are given by (4.37) with the reduced
matrix elements (determined from (3.14) (3.17) (3.70) and (3.71»
given by:
(j tJ'IIQlljtJ) = - [j(j :~)]1/2 c5~'_~" (4.52)
(/ - /)X 2Jl /2
<J - 1 tJ'IIQIUtJ) = - i [ j(2j _0 1) c5 nn" (4.53)
. ,.
<J + 1 tJ IIQIIJ tJ) = -I
.[«(j(j ++ 1)2 - j6)X 2Jl /2
1)(2j + 3) c5~~,. (4.54)
The tJ-dependence follows from (4.1) and (4.3). These reduced

matrix elements, together with (4.37) and the usual matrix elements
of J K [as given by (III.3.8'), (III.3.22), and (III.3.23)], constitute a
representation in (yt of the enveloping algebra g(E3) of the three-
dimensional Euclidean group extended by parity. Whenjo = 0, then
<JIIQIIJ> = 0 and the space is given by (4.17). Equation (4.54) shows
thatj can become arbitrarily large, i.e.,jl = 00.
19 A. Bohm and R. B. Teese, J. Math. Phys. 17,94 (1976).

V.4 Parity 201
The energy spectrum of the symmetrical top, i.e., the matrix elements of
the energy operator H in the basis Ii j3 '1 >of ~, is obtained from (4.49) using
(4.51):
spectrum H = - 1 ( j(j + 1) - IA - IB A2 ) (4.55)
2IB IA
(j = A, A + 1, A + 2, ... ), (4.55a)
where A = jo is the conventional notation used in molecular spectroscopy.
The angular momentum spectrum (4.55a) follows from (4.44) or (3.74).
It is also intuitively (classically) clear, as the value of total angular mo-
mentum j must always be larger than its component A in the direction of
Q. The component along the direction Q (axis of the molecule) can take
the two values ±A = jox/lxl (the - sign corresponding to the dashed
line in Figure 4.1). As
(4.56)
parity states are not eigenstates of the operator Q;1i that represents this
component. Thus in parity eigenstates the molecule does not have definite
values for the component of angular momentum along the molecular axis. 20
From (4.55) it follows that the energy levels of the symmetric top, in which
the component of angular momentum along the top axis is fixed, are the
same as those of the simple rotator except that there is a shift of magnitude
proportional to A2 , and except that levels withj < A are absent. Each energy
level has, in addition to the usual (2j + I)-fold degeneracy, a twofold de-
generacy because of the two possible values of '1 = ± 1. The energy diagram
of this kind of symmetric top for the case A = 1 is shown in Figure 4.2, with
that of the rotator for comparison. The + or - sign gives the values of the
20 This shows the inadequacy of the classical picture ofa dumbbell with an electron revolving
about its axis, on which Fig. 4.1 is based. In the quantum-mechanical state with definite parity
the electron is neither revolving clockwise (positive value for j . a) nor counterclockwise (negative
value for j . a) around the direction a. It is also not in a mixture of states with clockwise and
counterclockwise revolving electrons. But it is in a pure state which is incompatible with either
of these eigenstates of J . Q.
j j
===+
4----+ 4===±
3---- ===+
2----+ 2===±
1---- 1 +
o + (0) ----------
(a) (b)
Figure 4.2 Energy diagram (a) of the rotator and (b) of the symmetric top. For the
symmetric top A = 1 is assumed. The dashed level with} = 0 therefore does not occur.
[From G. Herzberg, (1966), Vol. 1 with permission.]
parity n. The levels with a given value of j and n = ±, which according to

(4.55) should be degenerate, are drawn slightly separated. The selection rule
/1j = ± 1, 0 has already been derived. Thus for these kinds of molecules, in
addition to the Rand P branches given by (III.5.19), there also appear in the
vibration-rotation bands transitions with /1j = 0, i.e., a series of lines given
by
(4.57)
(Here and also in the formula (III.5.l9) for the Rand P branches, Vo is not
the oscillator's wave number but
va c - [(IA - IB)/IA](Bn,A~,) - Bn"Afn"))'
But as A(n') and A(n") are constant, this last term can be absorbed in a suitably
redefined vo.) This series of lines is called the Q-branch. The energy-level
diagram and the transitions for the case that the upper rotational band has
the value A = 1 and the lower band the value A = 0 are depicted in Figure
4.3.
J'
---------------------------8
--~~------~----------~-6
I • , I II I U I I I I • I
, I I I "'101 1 I I I I
I I I I "11 " , I I I I
;;;-r:;=s=:;::;:;;;:; ~ ~
~~~~ISi', lSiit'll:" Il:" Il:" ,
I
J"
9
oL
1IIIIIImllllllili
Figure 4.3 Energy level diagram for a band with P, Q, and R branches. For the sake of
clarity, in the spectrogram below, the lines of the P and R branches which form a single
series are represented by longer lines than those of the Q branch. The separation of the
lines in the Q branch has been made somewhat too large in order that the lines might be
drawn separately. The convergence in the P and R branches is frequently much more
rapid than shown. [From Herzberg (1966), with permission.]
V.4 Parity 203
!J.j = 0 transitions are not observed in the infrared spectrum of diatomic

molecules. Therefore A = 0 for the electronic ground state of the diatomic
molecules, i.e., the motion of the electrons in the electronic ground states is
such that no angular momentum about the interatomic axis results, so the
diatomic molecule in the ground state is a dumbbell. However, all three
branches have been observed in transitions between different electronic
levels of diatomic molecules. Therefore in excited electronic states A is in
general different from zero but has a constant value as long as the electronic
state does not change. Thus in an excited electronic state, the diatomic
molecule is a dumbbell with a flywheel rotating around its axis. Figure 4.4(a)
shows an example of the energy levels and transitions between an electronic
state with the electronic angular-momentum component A = 0 and in-
e
trinsic parity IJ = + 1 L +), and an electronic state with A = 1 II). e
Figure 4.4(b) shows the same for an electronic state with A = 0 and IJ =
e
- 1 L -) and the 1 II state. The splitting between the two "degenerate"
levels of opposite parity, which was indicated in Figure 4.2, and which should
not be there according to (4.55), does in fact occur. It is a consequence of
the interaction between the electronic motion and the rotation of the molecule
(A-type doubling), which was not taken into account in the derivation of
(4.55).
The transition frequencies given by (4.57) are in the visible or ultraviolet
region, as Vo is in this region for transitions between different electronic
states.
J J
(a)+ --..--------~- c (a)+
00 d 4 (.)- ~ 4
'n : =t:=+:=:+:=;t::;::=+::t::
(s)
(a)+
~ ~
c 3
d (a)+
(a) +
(aj+ ~ 2
(s) • (.) -
I I
III j II ~ ~
(.) - =+=+=;:::F;:#J~RF=I~= c 1 (.)
(a)+ ~ d
(a) +
;;:;
I I ~~ I I
~~~~ ~ ?~~a:-
~
N ~
- ~
0 ~ ~
N
~
..,
~
... ~
~
~ ~ ~ ~ it" it" it" it"
~ I
(a)- ~-!-!-+--++++-!-!-!.........I- (.)+
I
(a)-
(a) - ---,---1--+--+- 44J-4-...L...-- (.)+
(.)+ _-.J.-4........-4.L...-.J._ _ __ (a)-
(a)-
00+ ===::lC====== 01 (.)+
(a)
I
o
(a) (b)
Figure 4.4 Energy-level diagram for the first lines of (a) a 1 n - 1 l: + transition and
(b) a 1 n - 1l:- transition. For the sake of clarity, the A-type doubling in the 1 n state
has been greatly exaggerated. The broken arrow to the left in (a) gives R(2)-Q(2), and
the one to the right gives Q(3)-P(3). Their difference gives the sum of the A doublings
for J = 2 and J = 3 in the upper state. [From Herzberg (1966), with permission.]
Problem
1. Determine the energy levels of the (isotropic) three-dimensional harmonic
oscillator and the degree of degeneracy of each energy level. Determine the possible
values of the orbital angular momentum in the nth energy level. To do this, proceed
as follows:
(a) Use the basic assumption IV to obtain the space of physical states and the
product basis from the combination of three one-dimensional oscillators. In
this product basis the energy operator H = p 2 /(2m) + (mw 2 /2)Q2 is diagonal.
Use this fact to obtain the spectrum of H and the degeneracy of its eigenvalues.
(b) Introduce the angular momentum operator L j = £ijk QjPk and the c.s.c.o.
H, L 2 , and L 3 . This c.s.c.o. is not diagonal in the product basis obtained in (a).
Introduce a basis o~eigenvectors of the c.s.c.o. H, U, and L3 (the angular
momentum basis), and determine the transition coefficients between the
product basis and the angular-momentum basis. Determine the possible
values of I for a given energy level En.
CHAPTER VI
Hydrogen Atom-
The Quantum-Mechanical Kepler
Problem
The classical Kepler problem is described in Section VI.2, where the Lenz
vector is introduced. Section VI.3 gives the algebra of angular momentum
and the Lenz vector, and Section VI.4 describes the representation of this
algebra. In Section VI.5 we present the algebraic derivation of the hydrogen
spectrum and discuss, at the end, fine-structure effects.
VI.l Introduction
An atom is described in everyday language as a nucleus with positive charge

+ Ze and Z electrons, each with negative charge - e, that move around the
nucleus in fixed, stable orbits. The simplest atom is the hydrogen atom, which
-in this language-consists of the proton as the nucleus and one electron
moving around it in the field ofthe Coulomb electrostatic force of the proton.
Thus the hydrogen atom is considered to be the realization of the quantum-
mechanical analog of the classical Kepler problem.
The quantum-mechanical Kepler problem was first solved by Pauli (in
1925) using the Heisenberg commutation relations. Later it became custom-
ary to treat the hydrogen atom as a solution of the Schrodinger equation
for the 1/r potential, which was introduced in 1926. Our presentation is based
on Pauli's and Bargmann's treatment. 1
To conjecture the algebra of observables for this problem we shall follow
our general procedure for quantum-mechanical systems that have classical
1 W. Pauli, Z. Phys. 36,336 (1926); V. Bargmann, Z. Phys. 99, 576 (\936).

205
206 VI Hydrogen Atom-The Quantum-Mechanical Kepler Problem
analogs and obtain the relations between the quantum-mechanical observ-

abIes by correspondence from the relations between the classical observables.
We therefore recall some facts about the classical Kepler problem.
VI.2 Classical Kepler Problem
Let Xi be the coordinates of the electron (with charge - e), and let Pi be its
momentum components. Let the proton (with charge +e) be located at
the origin of the coordinate system. Then the Coulomb force between
electron and proton is given by
where
n = xlr and r = Ixl = JLiXiXi,
and the potential energy (Coulomb potential) is given by

e2
VCr) = - - .
r
[Here the charge e, the system constant, is measured in electrostatic units:

e = 4.803 x 10- 10 esu; 1 esu = 1 cm dyn 1/2 = 1 gl/2 cm 3 / 2 sec- 1. If e is
measured in the practical unit of charge, e = 1.602 x 10- 19 coulomb, then
F is given by IFI = (4nfo)-1(e2Ir2) with (4nfo)-1 = 8.99 x 10 9 kgm 3
sec - 2 C - 2]. Therefore the energy of the electron is given by
p2 e2
E= - - - (2.1)
2m r'
wherem is the mass of the electron. [The electron mass me = 9.11 x 10- 31 kg
is much smaller than the proton mass mp = 1836me ; if one wants to take
into account that mp is not infinite, then min (2.1) is given by the reduced
mass
and p is not the momentum of the electron but rather
- the reduced momentum.]

Besides the energy E, there are two other constants of the motion: the
angular momentum
I=xxp, (2.2)
VI.2 Classical Kepler Problem 207
Figure 2.1
and the Lenz vector [or Runge-Lenz vector; cr. Barger and Olsson (1973)J
p x 1 e2
a= ---+-x.
m r
(2.3)
The closed orbits of the classical Kepler problem are ellipses (Figure 2.1).
1 is an axial vector that is perpendicular to the plane of the orbit. As 1 is a
constant of the motion, the orbit lies in some plane through O. For the
Kepler problem, V = - e2 /r, not only is the plane of the orbit fixed, but the
orbit itself is also fixed, i.e., the major axis PAis fixed and the ellipse does not
precess. The constant of motion that is used to characterize the orientation
of the major axis in the orbital plane is the Lenz vector a, which is directed
along the major axis from 0 to A.
The constant of motion ais a special feature of the 1/r potential. Whereas
) is a constant of the motion for every spherically symmetric system [i.e.,
for V(x) = V(I x DJ, a is a constant of the motion only for V(x) ex 1/r; a
small deviation of the potential from this form causes the axis P A of the
ellipse to precess slowly in the plane perpendicular to I, so that the orbit is
not closed.
For later comparison with the quantum-mechanical case, we list here
some relations between the classical constants of the motion. 2 Let a par-
ticular orbit have the semimajor axis with length a,
the semiminor axis
with length b, and eccentricity £ = J1 - b2 /a 2 • Then
e2
E = - 2a'
il 2 = 2E F + e4 = (e2£)2. (2.4)
m
One defines
a = Ii m/J -2mE.
From (2.3) follows immediately
). a = ). a = 0, (2.5)
and a short calculation using (2.4) shows that
m 2 e4
a 2+F-
- -2mE' (2.6)
2 See, e.g., Barger and Olsson (1973), Chapters 4-3.

VI.3 Quantum-Mechanical Kepler Problem
The quantum-mechanical observables of the quantum-mechanical Kepler

problem for the hydrogen atom are obtained by replacing the numbers
Pi' Xi in (2.1), (2.2), (2.3) with the Hermitian operators Pi and Qi that fulfill the
Heisenberg commutation relations (111.2.1).
For the angular momentum this has been done before; the angular
momentum operators are given by
(i = 1, 2, 3). (3.1)
The energy operator for the hydrogen atom is conjectured from (2.1) to be
H = ~~ - e2 Q- 1, (3.2)
where Q - 1 is the inverse of the operator Q = J Qi + Q~ + Q~.

[The mathematical meaning of Q- 1 requires some explanation.
As Qi are Hermitian, A = Ii Q? is a positive-definite operator. An
operator A is positive definite if (1, Af) 2 0 and if from (1, Af) = 0
follows! = 0 for all! E £: In particular, A has the property that its
spectrum {a} (all its eigenvalues) is nonnegative. For every
Hermitian positive-definite operator A, one uniquely defines a
positive-definite Hermitian operator B such that B2 = A. B is
written as B = JA. It has the same system of eigenvectors as A,
commutes with every operator that commutes with A, and has the
eigenvalues {Ja}. Often this definition of the operator is JA
inadequate, and one must allow for eigenvalues {-Ja}.
The inverse operator B- 1 of an operator B is the operator that
(if it exists) has the property that B- 1 B = BB- l = I. If B is a
Hermi tian operator with spectrum {b}, then the spectrum of B- 1 is
{lib}, and B- 1 has the same system of eigenvectors as B.]
To conjecture the quantum-mechanical Lenz vector from (2.3), we have

to agree first what the quantum-mechanical correspondence of the numbers
(p x l)i = fijkPj1k should be. As the operators P j and Lk do not commute,
the operator fijkPjLk is not Hermitian, but instead
(fijkPjLk)t = fijkLkPj = fijkPjLk - 2iliP i ·
As we want the quantum-mechanical observables Ai that correspond to the

components of the Lenz vector to be Hermitian operators, we replace the
classical quantity by the symmetrized product,
fijkPj1k ~ f ijk HPj, Ld,
where the anticommutator {A, B} of two operators A and B is defined by

{A, B} == AB + BA.
Thus the quantum-mechanical Lenz vector is

_ 1 e 2 Qi
Ai == 2m tikl{P" Ld + Q. (3.3)
For the sake of convenience we shall now introduce some new units. So
far the momentum P has been measured in eV/c, where c is the velocity of
light, and the position Q has been measured in cm. The canonical commuta-
tion relations read
where h=6.58 x 1O- 18 eVsec= 1.97 x lO- s eVcm/c with the velocity
of light c = 2.998 X 10 10 cm/sec. The Planck constant 21th was the con-
version factor between the measurement of energy E in units of eV and the
measurement of energy v in units of sec- 1 (frequency units): E = (21th)v = hv
(Planck - Einstein relation); or it is the conversion factor between the momen-
tum in units of inverse length (cm -l)-i.e., as the reciprocal of the de Broglie
wavelength, 1/A-and the momentum in standard units (erg sec/cm):
p = 21th/A (the de Broglie relation). After we have convinced ourselves
that frequency and energy or inverse wavelength and momentum are the
same physical quantities measured in diff<!rent units, it is no longer necessary
to carry the conversion factor h between these different units. 3 Therefore we
put Ii == 1 in all our calculations, which means that we replace the position
Q measured in cm with the position Q/Ii measured in (eV/c)-l. We call
Q/h simply Q. Thus, the canonical commutation relations now read
(3.4)
For the particular case of the electron in a hydrogen atom we simplify the
notation further by using meH as the energy operator, i.e., instead of p 2 /2m
- e 2 /Q we use the energy operator measured in units of eV . me, which we
simply call H:
p2 a
H=--- where (3.5)
2 Q'
The quantity
is the Sommerfeld fine-structure constant. In these units the Lenz vector is

given by
(3.6)
3 That energy and momentum are also the same physical quantities measured in different
units follows from relativity.
210 VI Hydrogen Atom~ The Quantum-Mechanical Kepler Problem
The commutation relation of the angular-momentum operators is given in

these units by
(3.7)
which is immediately seen from (111.2.14).
Now that we have conjectured the quantum-mechanical observables and
thus the algebra of operators for the quantum-mechanical hydrogen-atom
we want to derive the properties of this algebra and compare the results
with the experimental data. In particular, we would like to know the spec-
trum of H, which will give us the energy levels of the hydrogen atom. It
is clear that if we knew all algebras of operators in linear spaces generated
by Pi> Qj that fulfill (3.4), then the algebra of observables and space of states
for the hydrogen atom would be one of them. For the hydrogen atom we
need only one representation, i.e., one particular algebra of operators in one
particular space of states. Instead of giving the precise mathematical speci-
fications for this representation, we shall just go ahead and derive it. We
remark that the space of physical states (subspace of the Hilbert space) for the
hydrogen atom is different from the one for the three-dimensional oscillator,
though the basic commutation relations (3.4) are the same.
As in the case of the oscillator, the momentum Pi and the position Qi of
the electron are not physical observables, in the sense that the physical
system (hydrogen atom) cannot be prepared in approximate eigenstates of
Pi or Qi; such a preparation would break the hydrogen atom apart. The
hydrogen atom, like every nonrelativistic quantum-mechanical system,
appears in eigenstates (pure or mixtures) of the energy operator. The energy
operator (3.2), (3.5) of the hydrogen atom differs from that of the oscillator,
which is of the form p2 + Q2 [cf. (11.2.1b)]. The space of physical states
is the space spanned by the physically preparable states. For the oscillator
these are eigenstates of p2 + Q2, and for the hydrogen atom they are
eigenstates of H in (3.5). Since the energy operators differ, the spaces of
physical states differ.
In the previously considered cases of the oscillator and rotator, new
observables were introduced, as functions of the Pi and Qj, which were more
directly related to the physically preparable states (e.g., the Li for the rotator).
After the relations between these new observables had been derived as
consequences of (3.4), we could forget about the origin of these relations and
consider them as new fundamental (defining) relations for the particular
physical system under consideration. We shall follow the same procedure
here for the hydrogen atom.
In order to decide what to choose for these fundamental observables in
the case of the hydrogen atom, we recall that a state vector can be an eigen-
vector of two different operators only if these two operators commute.
Physical states for stationary nonrelativistic quantum-mechanical systems
appear to be in general eigenstates (or mixtures thereof) of the energy
operator. Therefore we choose operators that commute with H as these new
observables; these are the quantum-mechanical constants of the motion, as
we shall discuss later when we consider time development (Chapter XII).
As Ii and iii are the classical constants of the motion, we expect that the
angular-momentum operator Li and the Lenz operator Ai will be constants
of the motion, i.e., commute with H, and are therefore to be chosen as our
new fundamental observables for the hydrogen atom. A straightforward
calculation (Problem 1) shows that as a consequence of the Heisenberg
commutation relations (3.4) it follows that indeed
[H, LJ = 0 and [H, AJ = o. (3.8)
As H commutes with Li and Ai, it will commute with the whole algebra
generated by Li and Ai. Consequently there is no operator in this algebra
that transforms from a given eigenvector of H with eigenvalue E to a vector
that is not an eigenvector of H with the same eigenvalue E. Let us denote the
space of eigenvectors of H with eigenvalue E by ~(E). Then all L;, Ai and all
(well-defined) functions A = A(L, A) of them transform a vectorfE ~(E) into
a vector Af = 9 that is again in ~(E) with the same eigenvalue E. In other
words, the algebra of Lj, Ai leaves the eigenspaces ~(E) invariant. Let us
therefore first consider this algebra and investigate the structure of the
eigenspaces ~(E) of H.
To identify this algebra better (and to see that it is a simple, well-known
mathematical structure) we define a new set of operators:
(3.9)
We have to ask ourselves whether the definition (3.9) makes sense. On the
space ~(E) there is no problem, because on ~(E) J - 2H = J - 2E, i.e.,
J - 2H is a number, so on ~(E) Ai and Ai differ just by a constant factor.
Thus the restriction of Ai to ~(E), which we again call A;, is well defined.
Ai is also well defined wherever J- 2H is well defined. This is the case for
the space of vectors ffor which (J, ( - 2H)f) > 0, according to the definition
of the square root and in verse of an operator. This space is called the negati ve-
energy space, or space of bound states. We restrict ourselves to this space.
[For the space where + 2H is positive definite, instead of Ai one defines
A; = -J2ii Ai and finds that these A; fulfill commutation relations that
differ from those of Ai by a factor of ( -1), i.e., Ai and iA; fulfill the same
commutation relations; see beloW.]
We denote the negative-energy space by~. On~, Ai given by (3.9) is well
defined. From (3.8) it follows immediately that
[H, AJ = O. (3.10)
By a straightforward but lengthy calculation (Problem 1) one can show

that as a consequence of the Heisenberg commutation relation (3.4) the Ai
fulfill the following commutation relations with Li and with each other:
(3.11)
(3.12)
With the Li and Ai one defines the operators

C l == A2 + L2 + A~ + A~ + Li + L~ + LL
= Ai (3.13)
C2 == A·L = L'A = AILI + A2L2 + A3L3' (3.14)
These operators have the property that they commute with the Ai and L j (and
consequently with the algebra generated by Ai' L) if Ai and L j fulfill the
commutation relations (3.7), (3.11), (3.12):
[C l , LJ = 0, [C l , AJ = 0, (3.15)
[C 2 , LJ = 0, [C 2 , AJ = 0. (3.16)
The algebra generated by the Li and Ai that fulfill the commutation rela-
tions (3.7), (3.11), (3.12) is called is'(SO(4)).
[It is the enveloping algebra of the four-dimensional rotation group

SO(4). The algebra generated by Li and Ai = .j2iiAj, where 2H
is positive definite, is the enveloping algebra is'(SO(3,1)) of the
group SO(3, 1). cf. Appendix to Section V.3]
The operators C l and C 2 are the invariant or Casimir operators of is'(SO(4)).

In an irreducible representation space of is'(SO( 4», i.e., a space that is obtained
by applying every element of is'(SO(4» to one angular momentum eigenvector
(ladder representation), the operators C l and C 2 are multiples of the identity
operator. That is, they each have only one eigenvalue, Cl and C2 respectively,
and these values characterize the representation space [in the same way as
the value j(j + 1) characterizes the representation space of is'(SO(3))]. For
the particular case where Li and Ai are defined by (3.1), (3.6), (3.9), one can
calculate (Problem 1) that as a consequence of (3.4),
Cl = a2( -2H)-1 - I, (3.17)
C 2 = 0; (3.18)
i.e., the operator C I is related to the energy operator (3.5), and C 2 is the zero
operator. The relations (3.17) and (3.18) are the quantum-mechanical
analogue of the classical relations (2.6) and (2.5) between a i and Ii' Thus the
°
space ~(E) is a particular representation space of is'(SO(4)), namely the one
in which C2 = and C l = a 2 ( -2E)-1 - 1.
If we know all possible representation spaces of is'(SO(4)) that fulfill
(3.18), then we know the properties of all the operators L i , Ai given by (3.1),
(3.3), (3.9) in the spaces ~(E), for all possible values E with E < 0. It will
turn out that Cl cannot take any arbitrary real value, but rather that the
spectrum of C l is discrete. [For the invariant operator J2 of S(SO(3» we
found the same result, namely that only the discrete set of eigenvalues
j(j + 1) with j = 0, t, 1, ~, ... is possible.] Consequently from (3.17) it will
follow that the spectrum of the energy operator H is discrete and the spectrum
of H is obtained by (3.17) from the spectrum of C l'
VI.4 Properties of the Algebra of Angular Momentum and the Lenz Vector 213
Our task is therefore to find all the representation spaces of 6"(SO(4)),

in particular those that fulfill (3.18). This can be done in essentially the same
way as it was done for 6"(SO(3)) in Section 111.3, except that the calculations
for 6"(SO( 4)) are more involved. In the next section we shall give a description
of the properties of the representation spaces of 6"(SO(4)) which should
suffice for the understanding of the subsequent material. A derivation of the
representations of 6"(SO(4)) has been given in the Mathematical Appendix
to Section V.3.
VIA Properties of the Algebra of Angular Momentum and

the Lenz Vector4
We choose as basis vectors in ~(E) the angular-momentum vectors II m).
This is possible because
H, L2, L3
is a system of commuting operators. Whether this is a c.s.c.o. for the hydrogen
atom can only be decided by comparison with experiment. Therefore we
make the assumption that this is a c.s.c.o.-i.e., that I, m and the energy are
the only quantum numbers of the hydrogen atom-and we shall later see
that this is true to a high degree of accuracy (neglecting spin effects). Thus
II m) is a basis system in ~(E). Since A is a proper vector operator with
respect to the angular momentum, as stated by (3.11) and Up Ai Up = - Ai'
it follows from the Wigner-Eckart theorem that
A"llm) = 11- 1m + K)<lm 1 KII- 1 m + K)<I- 1I1AIII)
+ 11m + K)<lm 1 Kllm + K)</IIAIII)

+ 11+ 1 m + K)<lm 1 KII + 1 m + K)<I + 11IAIII). (4.1)
Since II m) has been assumed to be a complete basis system, there is no
parity doubling, and since Ai is a proper vector operator, it follows as in
(VA.25) that the reduced matrix element <IliA III) = o. [This can also be
seen to be a direct consequence of (3.18), which is related to the Up transfor-
mation property of A . L]. If we insert into (4.1) the explicit values of the
Clebsch-Gordan coefficients we obtain, as a consequence of (3.11) and (3.18),
the following for A" :
Aollm) = JI 2 - m2 C1 1I- 1 m) - J(I + I? - m2 C1 + 1 11 + 1 m),
(4.20)
A+11Im) = -J(I- m)(I- m - 1)/2C1II- 1m + I)
-J(I + m + 1)(1 + m + 2)/2C I + 111 + 1 m + 1),
A_ 111m) = -J(I + m)(I + m - 1)/2Ctll- 1 m - I)
-J(I- m + 1)(1- m + 2)/2C I + 111 + 1 m - I),
4 For a derivation see Appendix to Section Y.3.

214 VI Hydrogen Atom~ The Quantum-Mechanical Kepler Problem
where the C z are as yet undetermined functions of the discrete variable I,

related to <1- l1IAIII) by Equation (V.3.14).
The C z are determined by (3.12). We state the result: For every natural
number n = 1, 2, 3, ... , there is a function qn) such that (3.12) is fulfilled.
This function is
(n) _ . ~1
n - 2 C~) = o. (4.3)
Cz - I 4[2 _ 1 '
From (4.2) and (4.3) one finds the following properties: For every number
n = 1,2, 3, ... , there is an irreducible representation space ~(n). The lowest
angular momentum in ~(n) is 1 = 0, because starting from a given value 1
by application of Ak one can always reach the value 1 - 1, according to
(4.2) and (4.3), unless 1 = 0, because then Cbn ) = O. The highest angular
momentum is n - 1, because according to (4.2) one can always reach the
value I + 1 from a given value 1 by applying An unless Cl~\ = O~which,
however, is the case according to (4.3) for 1 = n - 1. As the II m) are ortho-
gonal for different values of I, the space ~(n) is the orthogonal direct sum of
the irreducible representation spaces ~I of angular momentum:
n- 1
.~(n) -------> I EB .~l, (4.4)
(;(SO(3» 1=0
where the symbol ~ is to indicate that the two spaces are the same when only
the algebra c&"(SO(3» acts in them.
The action of L 3 , L± in each of the spaces ~l and therefore in ~(n) has
been derived in Section II1.3. In particular,
Vl/m) = l(l + l)l/m). (4.5)
Thus Li and Ai, and consequently the whole algebra c&"(SO(4», is known in
each ~(n)(n = 1,2, 3, ... ). To indicate that the vector II m) is an element of
,~(n), i.e., that AK is given by (4.2) with the particular Cl n ), we label the vector
also by this number and write: I n I m). A straightforward calculation, using
(4.2) and (4.3), gives that the eigenvalue of the operator C 1 in the space
.~(n) is C 1 = n 2 - 1:
(4.6)
This gives the set of discrete numbers that are possible for C l'
The space ~(n) is not only an irreducible representation space of c&"(SO(4»,
but also of c&"(SO(4» extended by parity Up:
UpLiUP = + Li, (4.7)
UpAiUp = -Ai' (4.8)
Applying the considerations of Section V.4 to the space ~(n), where there is
no parity doubling, one has
Up In 1m) = (- 1Y1J n In 1m), (4.9)
where '1n = + 1 or '1n = -1, but may be different for different spaces 91(n).
We will study the n-dependence of '1 below.
VI.S The Hydrogen Spectrum
In the previous section we described all the irreducible representation spaces

91(n) of the algebra of angular momentum and the Lenz vector [i.e., all
irreducible representations of S(SO(4» that fulfill (3.18)]. The relation (3.17)
connects these spaces with the energy eigenspaces 91(E) of negative energy
E <0:
C1 = (n 2 - 1) = a 2 (-2E)-1_1. (5.1)
Thus the energy in each PA(n) is given by
a2
E = En = -2n 2 '
and the 91(n) are all possible energy eigenspaces of the hydrogen atom.
The space of physical states of the hydrogen atom PA is the direct sum of
all these PA(n)
00
91 = L Ef) 91(n), (5.2)

n= 1
and the spectrum of the energy operator in 91 is

spectrum H = En = -a 2 j(2n 2 ) (n = 1,2,3, ... ). (5.3)
n is called the principal quantum number.
[The meaning of infinite linear combinations of In E 91(n) (i.e.,

L:'= tin) requires a precise definition. It may mean infinite sum in
the Hilbert-space sense (L:'= 1 I In 112 < (0), in the Schwartz-space
sense (L:'= 1 IIInl1 2n2P < (0), or in some other sense. Even if (5.2) is
meant in the Hilbert-space sense, 91 is "different" from the Hilbert
space Yf = {L~ 0 4>. I¥p 2 + Q2)4>. = (V + J;)4>. and L~ 0 114>.11 2
< oo}, which is the usual Hilbert space of quantum mechanics. The
isometric map between 91 and Yf is physically meaningless, as it
would map elements that represent different physical states into
each other.
Physical states of the hydrogen atom are in any event only
vectors in the dense subspace Ln9l(n) of the Hilbert space, where
the sum extends over an arbitrarily large but finite number of n.
Any limit n ~ 00 is a mathematical idealization, and physically
for n ~ 00 the physical system ceases to be a hydrogen atom; it
ionizes and becomes a system consisting of a proton and electron.
One can, of course, consider the enlarged physical system, consisting
of hydrogen-atom states as well as e1ectron-and-proton states. For
such a physical system ~ is not the total space of physical states,

though this space may still be a dense subspace of yr. The space of
physical states of this system is the direct sum of ~ and a space fYl'
that is the continuous direct sum of irreducible-representation
spaces of the algebra 8(SO(3, 1)) generated by Li and Ai = Ai' Jill
The space ~ is an irreducible-representation space of the algebra
8(SO(4, 1)) [and also of 8(SO(4, 2)) and of 8(E 4 )]; the space fYl'
is an irreducible-representation space of 8(SO(3, 2)). Such algebras
have been called spectrum-generating algebras, the underlying
groups are called spectrum generating groups.]
Transitions between different energy levels are performed by the position

operator Qi (dipole transitions) and powers thereof (multi pole transitions).
Thus the Qi are observables that transform between different spaces fYl(n).
Using the property that the Q" (I( = 0, ± 1) are the components of a proper
vector operator, one obtains
Q"lnlm)=Iln'I-1m+I()<lm11(11-lm+I()(21+ 1)1/2qi"n
n'
+ I In' 1+ 1 m + 1()<lm 11(11 + 1 m + 1()(21 + 1)1/2q7'n,

n'
(5.4)
where the reduced matrix elements
q/,n = (21 + 1)-1/2<n' 1- 111Qlln I),
iJi"n = (21 + 1)-1/2<n' I + 111Qlln I),
depend upon n', n. They can be computed (Problem VII.1), but at present
we only need the property that they are different from zero for all n' and n.
The physical meaning of this statement is that there exist dipole transitions
between all energy levels of the hydrogen atom.
We can determine the n-dependence of the intrinsic parity rJn using the
Up transformation property of the Q". Applying Up to (5.4) gives, according
to (4.9),
UpQ"lnlm) = I(-lY- 1rJn,ln' 1- 1 m + I()<lm 11(11- 1 m + I()
n'
X (21 + 1)1/2q/,n
+I
n'
(- 1)1 + 1 rJ n' In' 1 + <
1 m + I() I m 1 I( II + 1 m + I()
X (21 + 1)1/2iJi"n. (5.5)
But use of (V.4.1) gives
UpQ"lnlm) = -Q"Uplnlm) = -(-1YrJnQ"lnlm). (5.6)
Inserting (5.4) into the right-hand side of (5.6) and comparing the co-
efficients with those of (5.5) (i.e., making use of the linear independence of
the basis vectors), one finds
(5.7)
As the reduced matrix elements are different from zero, q't n #-

it follows that
° and fi't n #- 0,
'1n' = '1n = '1, (5.8)

i.e., '1n is independent of n, and the parity of the energy eigenstate is
Up I n 1m) = ( - 1Y'11 n 1m), (5.9)
where '1 is either + 1 or -1. Equation (5.9) means that the parity of a state
with a definite angular-momentum value I is the same in all energy spaces
~(n). By convention one chooses '1 = + 1; only the relative parity has a
physical meaning.
Each energy eigenspace ~(n) of the hydrogen atom is, according to (4.4),
the direct sum of angular-momentum eigenspaces ~I; thus each energy level
En contains the n angular momenta
I = 0, 1, 2, ... , n - 1.
Each angular-momentum eigenspace ~I in Bl(n) is, according to (111.3.24),
the direct sum of the 21 + lone-dimensional spaces ~~ :
+l
BlI ~ L EB~~, (5.10)
8(SO(2» m=-I
which are spanned by each vector In 1m). 6"(SO(2» denotes the subalgebra
of 6"(SO(3» that is generated by L 3 • The dimension of Bl(n) is therefore
n-1
1=0
L (21 + 1) = n 2, (5.11)
i.e., there are n2 linearly independent eigenstates of H with eigenvalue En, or

2
the eigenvalue En is n -fold degenerate.
In order to compare the calculated energy spectrum with the experimental
data measured in eV, we insert a = m e e 2 jn from (3.5) into (5.3):
(5.12)
(recall that En is the energy in eV . me' E~ is the energy in eV). The number
R" = ~h~
4= C;e
2 (2)2
:n
can be calculated using
m e c 2 = 0.51 x 10 6 eV,
e2 = 23 x 10- 20 esu 2 = 23 x 10- 20 dyn cm 2 ,
e2 1
IX = cn = 137.036'
R" = me e4 j(2h2) is the Rydberg constant in energy units. Its numerical value
IS
R" = 13.6 eV. (5.13)

The energy levels of the hydrogen atom are plotted in Figure 5.1, where
details in the experimental energy spectrum have not been taken into account.
The numbers on the left give E~ - £'1' which is measured in eV. As one sees,
(5.12) is experimentally very well fulfilled. The energy levels are spaced closer
and closer to one another as n increases, and they finally converge to Eoo = 0
for n~oo. E>O was excluded by our restriction to the subspace where -2H
is positive definite; E > 0 is the region in which the quantum-mechanical
Kepler problem does not describe the hydrogen atom, but describes an
electron that is scattered by a proton (scattering states). Thus if we choose as
our physical system the hydrogen atom only, fjf of (5.2) is the space of physical
states. If we consider as our physical system the system that consists of the
eV
V,cm - 1
13
12
11
10
t.J 6 60000
I
"t 5
4
80000
3
100000
0L-~====================~~
"Oigure 5.1 Diagram of the energy levels of hydrogen.
hydrogen atom or electron and proton, then the space of physical states is
larger than ~, and ~ is the subspace of discrete energy eigenstates.5 E =
Eoo = 0 is the energy value at which this physical system ceases to be a
hydrogen atom; beyond this point the electron energy can take any value and
one can also prepare approximate momentum eigenstates of the electron. This
means that if one starts with hydrogen atoms in the ground state, one can
prepare the physical system in an approximate momentum eigenstate of the
electron if one transfers to it an energy of at least - E I, and that the system is
no longer a hydrogen atom. The continuous energy states correspond to the
hyperbolic orbits of the classical Kepler problem.
The numbers on the right-hand side of the diagram give the (negative)
energy in units of cm -I, i.e., the wave numbers Vn that are connected with
E~:
E'n
Vn = 2nhc' (5.14)
Inserting (5.12) gives for the wave numbers

e4me 1
vn = - h34nc n2' (5.15)
The quantities
R' = _l_R" = e4me = 3.29 x lOis sec-I, (5.16)

2nh 4nh 3
1
R = -- R
"
= -e4me
-3- =
5
1.097 x 10 cm-
1
(5.16')
2nhc 4nh c
are the Rydberg constant in frequency and wave-number units respectively.
(Originally it was R that was called the Rydberg constant.)
The lines connecting the different energy levels depict the transitions in
the emission or absorption of electromagnetic radiation. According to
(111.1.1), the frequency and wave number of the absorbed or emitted light
quanta are
V~m = IR'(:2 - ~2)1 (in sec-I), (5.17)
vnm = IR(:2 - ~2)1 (incm- I). (5.17')
5 It should be remarked that we are discussing the electron states only, i.e., describing the
electron under the external influence of the nucleus. If we wanted to describe the combined
system of electron and nucleus (proton), we would have to take as the space of physical states
the direct-product space of the electron space f1,i and the proton space J'fp : f1,i @ J'fp (axiom IV),
or f1,i @ J'fC. M., where J'fC.M. is the space for the center-of-mass motion. Neglecting the proton
structure (e.g., proton spin), J'fp is one-dimensional for protons at rest.
The number on each transition line in Figure 5.1 gives the wavelength of
the light in angstroms:
A = l/v nm = e/v~m,
with 1 A = 10- 8 cm = 10- 4 /1.

All transitions terminating in the same lower state form a spectral series
and each series of the hydrogen atom has a historical name. The series of
lines corresponding to transitions to the level n = 1 is called the Lyman
series. The frequencies of the Lyman series are
,
vim = R
,(11- 1)
m2 (m = 2,3,4, ... ). (5.18)
The line with the longest wavelength is e/v'12 = 1215.68 A. This line lies in
the ultraviolet region (cf. Figure 111.1.2). The series with n = 2,
(m = 3, 4, 5, ... ), (5.19)
is called the Balmer series. It lies in the visible region of the spectrum, and
(5.19) was first obtained empirically by Balmer in 1885. It played an excep-
tionally important role in the conjecture of the quantum theory of atoms.
The other series in Figure 5.1 with n = 3 (Paschen), n = 4 (Brackett), and
n = 5 (Pfund) lie in the infrared region.
To each energy level with energy value En corresponds the energy eigen-
space 9f(n) of dimension n 2 • The energy value En is n 2 -fold degenerate. Thus
if the energy of the hydrogen atom is measured as En, then the state is in
general a mixture (except for n = 1) and the (normalized) statistical operator
of this state is
(5.20)
where An is the projection operator on the n2 -dimensional subspace ~(n).
Energy measurements on the hydrogen atom do not reveal the value of the
angular momentum; they just give an upper limit to the possible angular
momentum values 1 = 0, 1, ... , n - 1.
The n 2 -fold degeneracy of each energy level is a peculiarity of the hydrogen
atom, or the l/r potential. This is connected with the existence of the constant
of motion Ai in addition to the angular momenta L i . In general one has only
a (21 + I)-fold degeneracy of the energy values as a consequence of [Li' H]
= 0.
However, the degeneracy for the hydrogen atom is only approximate,
and Figure 5.1 is only a rough picture of the experimental situation. The
above theoretical description, like every theoretical model, gives a description
of the situation in nature only up to a certain accuracy. Finer details of the
hydrogen spectrum are shown in Figure 5.2 for the low-lying energy values.
VL5 The Hydrogen Spectrum 221
Energy
=====
======--L
J 3D~1 2
J=========
3pll2 - - - 3D l l 2
3S
I
/2
3p Il2 = = = ==
L splitting due to
Lamb shifl
====::::;:
2S112 t-::========.------r--
t-
•
1 . Fine structure
2pll2
Lamb Shlfl
2pl 12 =='===
triplel
ISI /2 ~======~= Hyperfine spliuing
singlel
Figure 5.2 Low-lying energy levels of atomic hydrogen. (Not to scale. The levels with
different principal quantum number n are much further separated as is shown in Figure
5.1. This is indicated here by the broken lines separating energy levels that are far apart.)
The first number by each energy level gives the value of n. The symbol S, P, D
stands for the angular-momentum value I = 0, 1, 2, respectively (e.g., 2P
means n = 2, I = 1, and m can be anything such that -l:s m :S + I).
According to (5.12) all 2S and 2P states should have the same energy, as
shown in Figure 5.1. This is only approximately the case. The 2P states
split into two subsets 2p 3/2 and 2p l /2 with
E(2p3/2) - E(2pl/2) = 0.453 X 10- 4 eV = 1095 x 10 7 sec-I. (5.21)
Comparing this with the numbers of (5.12), (5.13), we see that our model
indeed gives an excellent approximation to the hydrogen spectrum; the
deviation (5.21) is only a fraction of one percent. The splitting (5.21) (fine
structure) and the 3p 3/2 to 3p l /2 and 3D 5/2 to 3D 3/2 splittings can be explained
if a new observable, the electron spin, is introduced. The superscripts 3/ 2 , 1/2,
etc. are connected with this new quantum number, the spin (see Section IX.4).
The splitting between the levels 2S 1/2 and 2p l /2 cannot be explained
even after the spin has been introduced. It is again an order of magnitude
smaller than (5.21):
E(2S 1/2 ) - E(2p1 /2) = 0.0353 cm - I
= 0.437 x 10- 5 eV
= 1058 x 10 6 sec-I
and is called the Lamb shift. The Lamb shift is a consequence of the inter-
action of the electron with the fluctuations of the radiation field, and its
explanation requires a new theory called quantum electrodynamics.
The splitting of each term IS I / 2 , 2S I / 2 , 2pl/2, etc. into two levels, the
hyperfine splitting, is another order of magnitude smaller than the Lamb
shift :
E(2SI/2 upper) - E(2SI/2 lower) = 0.0059 cm-1,
E(2PI/2 upper) - E(2PI/2 lower) = 0.0020 cm -I.
In order to explain this splitting one needs to include observables that
describe the internal structure of the nucleus (in particular, nucleon spin).
Summarizing, we have seen that our simple theoretical model describes
the features of the hydrogen spectrum very well. However, to explain the
fine details requires the introduction of new observables and even completely
new theories (quantum electrodynamics and nuclear physics). Any given
theory has only a limited domain of applicability, and this example of the
hydrogen spectrum serves very well to illustrate that behind each new digit
in an experimental number may be hidden a completely new domain of
physics.
Problem
1. Fill in the missing steps in the treatment of the hydrogen atom: Let Li be the com-
ponents (3.1) of the orbital-angular-momentum operator, and let H [as given by (VI.3.5)]
be the energy operator ofthe hydrogen atom. Let Ai denote the Lenz vector given by (3.6).
(a) Show that as a consequence of the canonical commutation relations the angular
momentum and the Lenz vector are constants of the motion, i.e., satisfy (3.8):
[Li> H] = 0 and [Ai, H] = o.
(b) Define Ai by (3.9). Show that Ai is Hermitian and a constant of the motion:
and
[Ai' H] = O. (3.10)
(c) Show that Li and Aj fulfill the commutation relations of (3.11) and (3.12):
[Ai' L j] = itijkAk>
[Ai' Aj] = itijkLk.
(d) Define the Casimir operators C 1 and C 2 by Eqs. (3.13) and (3.14). Derive (3.17)
and (3.18):
C1 = a2 ( -2H)-1 - I,
C2 = O.
The following fact is useful in carrying out the above derivations: If two operators A
and B commute, [A, B] = 0, then [f(A), B] = 0 for any well-defined function of the
operator A [cf. Equation (II.4.46)J.
CHAPTER VII
Alkali Atoms and the Schrodinger

Equation of One-Electron Atoms
In Section VII.1 the concept of perturbation theory is explained using the

example of the alkali atoms. Section VII.2 represents an algebraic calculation
of the matrix elements of Q-v (v = 1,3,4, ... ); the results are used for the
calculation of the energy values of the alkali atoms. Section VII.3 gives a
brief description of the solution of the Schrodinger equation for the hydrogen
atom, which is used for an alternative computation of the matrix elements of
Q- v and an evaluation of the alkali energy values. It also lists some properties
of the spherical harmonics which are used in the second part of this book.
VII.1 The Alkali Hamiltonian and Perturbation Theory

The alkali spectra are very similar to the hydrogen spectrum. This is suggested
by their classical model, according to which the alkali atom consists of one
electron - the" valence electron" - that moves in the Coulomb field of the
nucleus and in the average field of the other electrons that are in the orbits
closer to the nucleus.
The classical potential VA(r) for this" outer" electron has, for large values
of r, the form
VA(r) :::::: -e 2 Ir,
because the Z - 1 electrons in the inner orbits screen the charge of the nucleus
Ze, so that the effective charge will be e. For very small r (smaller than the
radius of the inner orbits)
223
224 VII Alkali Atoms and the Schrodinger Equation of One-Electron Atoms
Therefore VA(r) can be written

eZ
VA(r) = - - + VCr), (Ll)
r
where VCr) is negative and different from zero only in the neighborhood of
the nucleus. The classical Hamiltonian function of the outer electron is then
h = hHydr + VCr), (1.2)
where VCr) is a small perturbation of the Hamiltonian hHydr .
The exact expression for VCr) is not known since it depends upon the un-
known charge distribution of the core, which consists of the nucleus and
the inner electrons. If VA(r) is spherically symmetric, then VA(r) may be
expanded in the form
(1.2')
whereCo e is the charge, c 1 e = ed . xlr is the dipole contribution, C z e =

e%Xix)rZ is the quadrupole contribution, etc., of the core. For the alkali
atoms the extension of the core is small compared to the distance of the outer
electron, so that c 1 is small and contributions by the quadrupole moment
% and by higher multipole moments are negligible.
The Hamiltonian operator for the quantum-mechanical system is therefore,
in correspondence to (1.2),
H=K+V, (1.3)
where K is the Hamiltonian operator of the hydrogen atom,
pZ a
K=--- (1.3')
2 Q'
and
(1.3")
is the "small perturbation" of K. In (1.3), (1.3'), and (1.3") we have used the
units adopted in Section VI.3, i.e., we have made the replacements Q;/h -+ Qi
and meH -+ H.
The problem is now to find the spectrum of the Hamiltonian operator
H and its expectation values in the physical states. This can no longer be
done exactly. Also, we do not know the exact form of VCr) or V(Q), though
we know some of its properties. We can only hope, therefore, that the
spectrum of H will not differ significantly from the spectrum of K, i.e., that V
causes only a small perturbation of the spectrum of K.
The determination of the alkali spectrum is a problem of perturbation
theory, which will be discussed in more general terms in the next chapter.
One major factor in this kind of problem is the choice of an appropriate basis.
VII.1 The Alkali Hamiltonian and Perturbation Theory 225
One starts with eigenstates of K, i.e., chooses a c.s.c.o. that contains K. If

the other members of the c.s.c.o. do not commute with V, then Vapplied to
this eigenvector will not only change the eigenvalue of K but also the eigen-
values of these other observables. Thus Vwill perturb not only the spectrum
of K but also the spectrum of the other members of the c.s.c.o. Therefore
the c.s.c.o. should be chosen so that as many of their members as possible
commute with V.
For the alkali atoms we have
(1.4)
because
Consequently, we have the ideal situation where H as well as K can be

diagonalized together with L2 and L 3 • Thus the basis In 1m) of the hydrogen-
atom problem is a very appropriate basis to start with.
We again define
(1.5)
and we again have
Ct/nlm) = (A2 + V)lnlm) = (n 2 - 1)lnlm). (1.6)
Unlike the case of the hydrogen atom, however, it will not be true that the
full Hamiltonian operator H commutes with Ai' A2 , and C 1 :
[H, AJ = [V, AJ -=I- 0,
(1.7)
[H, C 1 ] = [V, C 1 ] -=I- 0.
Thus Hand C 1 (and consequently Hand K) cannot be diagonalized together,
i.e., there are no vectors in PIt that are eigenstates of H and also eigenstates
of C 1. Let us denote the eigenstates of H by IAim) and the eigenvalues
by E A :
H/Alm) = EAIAlm). (1.8)

H can be diagonalized together with L2, L3 because of (1.4).]
Now the question arises: What are the physical states, i.e., what are the
states that the alkali atoms can be prepared in? According to the basic as-
sumption III, the alkali atoms should be in an energy eigenstate (or a mixture
of energy eigenstates) if an energy measurement has been performed on the
system, i.e., the statistical operator should be
WH = AA' WW = AAI> or w~m) = AAZm (1.9)
if only the energy, or only the energy and angular momentum, or the energy,
angular momentum, and z-component of the angular momentum have been
measured, respectively. Here AA is the projector on the space of eigenvectors
of H with eigenvalue E A' AAZ projects on the space spanned by IAim>
(m = -I, -I + 1, ... , + /), and AAtm projects on the space spanned by

IA 1m).
If the state of the alkali atom has not been prepared by an energy measure-
ment but by a measurement of the observable C 1 or K, then, according to
the basic assumption III, this physical system should be in an eigenstate of
C 1, i.e., the statistical operator should be
or (1.10)
where An is the projection operator on the space of eigenvectors of C 1 with

eigenvalue n2 - 1.
Although at this stage we cannot exclude the possibility that states of
atoms are prepared by a C 1 measurement, this possibility appears very
unlikely, as C 1 and K have only an auxiliary meaning (except for the hydrogen
atom) and the physical observable is H. (Also, in Chapter XII we shall see
that states that do not change in time-and the states that correspond to
the energy levels ofthe atoms are likely to be well described by this property-
have to be eigenstates ofthe energy operator.) Thus, the occurrence of physical
states given by (1.10) would make the formulation of the theory very un-
satisfactory and would cause doubts about the appropriateness of III.
Therefore it will be interesting to see how the predictions of (1.9) and (LlO)
differ. In this chapter we shall perform the calculation for the alkali atoms
and see that an answer to the above question cannot be obtained from the
alkali spectra. In the later chapter on two-electron atoms (helium atom) we
shall, however, see that experimental data require (1.9), which reassures us
of the correctness of our basic assumption III-and the appropriateness of
the meaning of stationary states (cf. Chapter XII).
The value predicted for the measurement of the energy operator for the
state w~m) of (LlO) is, according to the basic assumption II,
(H)wUm1 = Tr(W~m)H) = Tr(AnlmH) = Tr(AnlmK)

K
+ Tr(Anlm V(Q»
= (nlmIKlnlm) + (n ImlV(Q)ln 1m)
a2
= - 2n 2 + (nlmlV(Q)lnlm)
a2
= - 2n 2 + fen, I) == E(n, I). (1.11)
For the case where the state is w~m) of (1.9), we have
(H)w'j.m) = Tr(w~m)H) = Tr(A;'lmH) = (A 1mlHIA 1m) = EAt. (1.12)
The term fen, l) in (1.11) is the matrix element of the small perturbing
Hamiltonian V:
fen, l) = (n 1ml V(Q)ln 1m). (1.13)

VII.2 Calculation of the Matrix Elements of the Operator Q - v 227
As In 1m) is known, this matrix element can be computed if V(Q) is known.

The value (n, l) does not depend upon m, because of (104). Therefore, for the
expectation value in the state Ani one obtains
Tr(W~H) 1 +1
<H)w<;2= T
r
W(l)
K
=-211L <nlmIHlnlm)=E(n,l).
+ m=-l
(1.14)
In contrast to the expectation value E(n, I), EAI is an element of the spectrum
of H. The spectrum of H and the states IA I m) need to be determined. The
spectrum of K and the states In 1m) are known from the hydrogen-atom
problem, and in order to calculate E(n, l) it only remains to calculate the
"small" matrix elements (n, I).
The eigenvalues of H and the eigenvectors are determined by a perturba-
tion-theoretical calculation. It will turn out that E(n, I) of (1.11) will be the
first-order approximation EW for the perturbation-theoretical calculations
of the eigenvalue E A/ • This is the reason we cannot distinguish between
(1.9) and (1.10). As E(n, l) cannot be calculated exactly [because V(Q) is
not known precisely], we can never tell from comparison with the experi-
mental energy spectrum of the alkali atoms whether E(n, I) is the exact
energy value or only a first approximation.
VII.2 Calculation of the Matrix Elements of the Operator Q - v
To obtain the correction term (n, I) = <n I ml V(Q)ln I m), we calculate

the matrix elements of the operators Q- v for any integer v. These matrix
elements will also be used for the evaluation of the fine structure splitting in
Section IXo4 below.
We proceed in the following way. We first derive some relations between
the operators Q-v, L2, and
(2.1)
The matrix element of one of these relations will give a recursion relation
for the matrix elements <n I ml Q-vi n 1m).
We introduce the operator
(2.2)
As a consequence of the Heisenberg commutation relation it follows that

(2.3)
[Equation (2.3) follows from

by use of
[Q, PJ = iQJQ. (2.4)
The proof of (2.4) proceeds as follows:
[Q2, PJ = [QjQj' PJ = QJQj' PJ + [Qj' PJQj = 2iQi'
On the other hand
[Q2, PJ = [QQ, PJ = Q[Q, PJ + [Q, PJQ = {Q, [Q, Pi]},
and consequently
{Q, [Q, PJ} = 2iQi' (2.5)
Furthermore,
[Q2, [Q, PJ] = (QQ[Q, PJ + Q[Q, PJQ)
- (Q[Q, PJQ + [Q, PJQ2)
= [Q, Q[Q, PJ] + [Q, [Q, PJQ]
= [Q, {Q, [Q, PJ}] = [Q,2iQJ = 0,
where (2.5) has been used. However, if Q2 commutes with [Q, PJ,
then Q = (Q2)1/2 commutes with [Q, PJ also (by the definition of
the square root of an operator). Therefore it follows from (2.5) that
2Q[Q, PJ = 2iQi'
which in turn gives (2.4).]
As a consequence of the definition ofL, Li = fijkQjPb and the definition

(2.2), it follows by a straightforward calculation that
1 2
2
P = P,
2
+ Q2 L . (2.6)
With this the energy operator K can be written

p2 L2 a
K=~+---. (2.7)
2 2Q2 Q
P; /2 is the operator of the kinetic energy of radial motion, and P r' the
conjugate to the radius operator Q = JQiQi' is often called the radial
momentum operator.
The radial momentum operator fulfills the commutation relation
(2.8)
[We prove (2.8) by induction on v. First note that

VII.2 Calculation of the Matrix Elements of the Operator Q-v 229
Then use of (2.3) gives [Q-1, PrJ = - iQ- 2. Consequently (2.8)

is true for v = 0, ± 1. Assuming (2.8) is true for v = n, one calculates
[Q-n-1, PrJ = Q-1[Q-n, PrJ + [Q-1, Pr]Q-n
= Q-1( -in)Q-(n+ 1) _ iQ-n-2 = -i(n + 1)Q-(n+2).
Thus if (2.8) is true for v = n, it is true for v = n + 1. In the same
way, one shows (2.8) is true for v = -en
+ 1) given that it is true
for v = -no Therefore (2.8) is generally true.]
As a consequence of (2.8) and (2.7) it follows by a straightforward calculation

that
[K, Q-1(V-1)] = t(v - 1)(2Q- viPr - VQ-(V+1». (2.9)
A special case of (2.9) is

[K, Q] = - iP r • (2.9a)
The relation
(2.10)
can be checked by inserting (2.7) into the left-hand side and using (2.8).
With (2.9) and (2.10) one calculates
[K, [K, Q-(v-1)]] = (v - 1)[ -v(v + 1)Q-(v+2)iPr - vQ-(v+1)p;
+ iv(v + l)(v + 2)Q-.<v+3) + L 2 Q-(v+3) _ aQ-(v+2)].
(2.11 )
Inserting (2.9a) into (2.9) and taking the commutator of the result with K
gIves
[Q-VCQ, K], K] - tv[Q-(v+1), K] + v[Q-(v+1), K]
= v ~ 1 [[K, Q-(v-1)], K] + v[Q-(v+ 1), K], (2.12)
where one term has been added to both sides.

Inserting (2.11), (2.7), and (2.9) into the right-hand side of (2.12) gives
[Q-V[Q, K], K] + tv[Q-(v+ 1), K] = 2vQ-(v+ l)K - (v + 1)Q-(v+3)L2
+ (2v + l)aQ-(v+2) + iv(v + l)(v + 2)Q-(v+3). (2.13)
The recursion relation for the matrix elements of Q-v is obtained if one takes
the matrix elements of (2.13) between the vectors In I m), for which
a2
Klnlm) = - 2n2Inlm),
Ulnlm) = l(l + l)lnlm),

L3lnlm) = minim).
230 VII Alkali Atoms and the Schr6dinger Equation of One-Electron Atoms
The left-hand side of (2.13) between the vectors 1n I m) gives zero, so that we
get as a recursion relation
a2
0= -v 2 <n ImIQ-(V+l)ln 1m) + (2v + l)a<n ImIQ-(v+2)ln 1m).
n
+ [-(v + 1)1(1 + 1) +iv(v + 1)(v + 2)]<nlmIQ-(v+3)lnlm).
(2.14)
If one takes the matrix elements of (2.13) between states 1 n I m) and 1 n' l'm),
one obtains recursion relations for the off-diagonal matrix elements. 1
For v = -1, Equation (2.14) gives
<nlmIQ-1Inlm) = a/n 2. (2.15)
The matrix elements of Q- 2 cannot be obtained from (2.14); they may,
however, be calculated using the wavefunctions found in the next section
and performing the integral (3.29) given there with VCr) = 1/r2. The result is
(2.16)
For v = -2, Equation (2.14) gives
n2 (
<nlmIQ1nlm)=2a 3-
1(1 +
n2
1») .
Of all the matrix elements of Q-v that can be obtained from (2.14), we list
the lowest ones:
a
<n I m 1Q- 31 n 1m) = 1(1 + 1) <n I m 1Q- 21 n 1m)
(2.17)
= n3/(1 + 1)(1 + 1)'
a 3n 2 - l(l + 1)
<nlmIQ- 4 Inlm) = n2 21(/ + 1) _ t <nlmIQ- 3Inlm). (2.18)
The constant a has, in our units, the dimension of an inverse length.

According to (VI.3.5),
(2.19)
The above Qis, because of our choice of units, not measured in cm but rather
in (eV/c)-l [cf. the discussion preceding Equation (VI.3.4)], i.e., the above Q
is in fact Q/h. If Qis measured in its usual units of cm, then the above a is to be
1 L. C. Biedenharn and N. V. Swarny, J. Math. Phys. 11, 1165 (1970).

VII.2 Calculation of the Matrix Elements of the Operator Q-v 231
replaced by a/h. Thus for Q measured in cm the above equations hold with
a replaced by
me ccx 1 1
(2.20)
f
a =--
h rB 0.529 x 10 8 cm .
rB is called the Bohr radius.

We can now give the energy value
a2
E(n,l) = - 2n 2 + <nlmlV(Q)lnlm) (2.21)
for any V that is a sum of positive and negative powers of Q. Equation

(2.21) is-as we mentioned above and as we shall see in the next chapter-
also equal to the first approximation of the perturbation expansion. We shall
use for V(Q) only the dipole term
V(Q) = -c 1 a/Q2, (2.22)
which -as was mentioned above-should be a sufficiently good approxima-
tion.
From (2.16) and (2.22) one obtains for E(n, l)
E(n, l) = - ;:2 (1 + n(~c~at))· (2.23)
Thus the representation spaces ~(n) of C(SO(4)) no longer have only one
energy value, for H does not commute with Ai' [In group-theoretical language
this is expressed as "V(Q) breaks the SO(4) symmetry."]
It is customary to write (2.23) in the form
22 a2
E(n, l) = - 2n*2 = - 2(n _ oY' (2.24)
where n* is called the effective quantum number and a = a(n, l) is called the
quantum defect. Comparing (2.24) with (2.23) and using the fact that C 1 is
small compared to the Bohr radius l/a, one obtains that the quantum defect
for the alkali atoms is given by
C1 a
a=-1-1 ' (2.25)
+2
The energy levels for lithium, sodium, and cesium are given in the diagrams
of Figures 2.1, 2.2, and 2.3, respectively. The energy levels are labeled in the
standard notation. The first number stands for the principal quantum number
n, and the letters s, p, d, f, ... denote the angular-momentum quantum-
number values 1 = 0, 1, 2, 3, ... , respectively; e.g., 3p means n = 3, 1 = 1.
The lines connecting different energy levels indicate dipole transitions. The
numbers in the lines give the wavelengths in angstroms. These transitions
fulfill the selection rule ~I = ± 1 [cf. (V.4.33)]. Thick lines indicate transitions
of higher intensity.
Comparing these energy diagrams with the energy diagram of Figure
VI.S.l for hydrogen, we observe the I-dependence of the correction term in
(2.23), which leads to a splitting ofthe levels with the same principal quantum
number. Thus (2.23) gives a good quantitative approximation to the experi-
mental energy spectrum. We observe further that some of the energy levels
split in contradiction to (2.23); e.g., there are two energy levels 3Pl and 3P2
of sodium, whereas according to (2.23) there should be only one. According
to (2.14) through (2.18), even if one takes for V(Q) the general spherically
symmetric expansion (1.3/1), the energy values should depend upon n and I
2S1/2 20 512 3/2 2F7/25/2

Voll s d f
5.37
5 IIcm-l
5~000
4f
4$
10000
4
15000
3$
3
20000
25000
2
1.84
30000
35000
7
40000
o 2$--~U-----J
45000
Figure 2.1 Energy levels of lithium atoms.
VIl.2 Calculation of the Matrix Elements of the Operator Q-v 233
2 2
Voll PII2 ~712 512
5.12
5.0 IIcm-1
e~
7s
65 5~
4f
5s
4.0
10000
45 15000
3.0
20000
2.1
20 25000
30000
1.0
35000
40000
o 3s-.L...L...L..----...J
4
Figure 2.2 Energy levels of sodium atoms.
only. The splitting of these levels with the same value of the quantum numbers
n, I will be explained by the introduction of the new observable spin, which
will be discussed in Chapter IX.
We also observe in the experimental energy diagrams of Figures 2.1-2.3
that for lithium there is no energy level with n = 1, that for sodium there are
no energy levels with n = 1 or 2, that for potassium (which is not depicted
here) there are no energy levels with n = 1,2, or 3, and that for cesium there
are no energy levels with n = 1, 2, 3,4, or 5. This means that our physical
system, which is the outer electron in the electrostatic potential VA(r) =
- e2 jr + V(r), cannot be in a state with principal quantum number n = 1
(lithium), n = 1 or 2 (sodium), etc. In Chapter X we shall see that this can
be explained as a consequence of a new basic assumption of quantum mech-
anics, the Pauli exclusion principle.
2
Volt $112
3.87 t----;;s--,_-.....t---r-----;;""--+----i,.....~
em l
9s
3.0 85
75
2.0
1.45
1.38
1.0
25000
30000
35000
40000
45000
Figure 2.3 Energy levels of cesium atoms.
VII.3 Wave Functions and Schrodinger Equation of the

Hydrogen Atom and the Alkali Atoms
The solution of the Schr6dinger equation for the hydrogen atom is so
familiar that it need not be discussed in detail here. It does not lead to any
new observable results, which have not already been discussed in the previous
chapter. The new quantities that it provides are the wave functions l/Inlm(x) =
<Xl, X 2 , x3lnlm) = <xlnlm). The quantity Il/I(x)1 2 represents, according
to the results in Chapters I and II, the probability density for the observation
of the electron in the state In I m) at the position x. Such a position measure-
ment, however, cannot be performed. Nevertheless, solving the Schrodinger
equation is the most common method of calculating the energy values,
and it is often a very convenient method. Furthermore, the wave functions
will be of physical importance for scattering problems. We give here only a
VII.3 Wave Functions of the Hydrogen Atom 235
brief description of this method, since detailed calculations are given in

many good standard textbooks on quantum mechanics.
The Schr6dinger equation for the one-electron atom is obtained by taking
the" scalar product" of the energy eigenvalue equation
Hlnlm) = Elnlm)
with the generalized eigenvector Ix) = IXtX2X3). Using C2.7) one obtains
then:
<xl (7 + 2~2 + VACQ») In 1m) = E<xln 1m). (3.1)
For the hydrogen atom VACQ) = VHydrCQ) = -a/Q. Equation C3.1) also
applies to any other one-particle problem in three-dimensional space if
VACQ) is replaced by the appropriate potential-energy operator, and the
In 1m) by the appropriate basis vectors.
In the basis of generalized eigenvectors Ix) ofthe Qi' the position operators
Qi act like multiplication by Xi :
(3.2a)
The momentum operators Pi' according to (11.7.53), act like differentiation
with respect to Xi :2
(3.2b)
We label the generalized position eigenvectors Ix) by the polar coordinates

(r, 8, ¢) instead of the Cartesian coordinates:
Xl = r sin 8 cos ¢, X2 = r sin 8 sin ¢, X3 = r cos 8. (3.3)
From (2.2) and (3.2) it then follows that
<xIPrll/!) = i1 or (0 + -;:1) <x II/!),

0
<x I P 2II/!) =
r
10(2
- r2 <
or r or x II/! ) ) . (3.4)
2 Note that the derivation of (II. 7.53) required, in addition to the Heisenberg commutation
relation, the assumption that p2 + Q2 had at least one eigenvector (or equivalently p2 + Q2
was essentially self-adjoint). Thus (3.2b) is not a consequence of (VI.3.4) only, but requires an
additional assumption. This assumption is natural for the case of the three-dimensional oscil-
lator, because then p2 + Q2 represents the observable energy. For the hydrogen atom the
energy is given by (2.1), and p2 + Q2 need not correspond to any observable for the hydrogen
atom. Thus (3.2b) is an additional assumption, and the description of the hydrogen atom given in
Chapter VI is not completely equivalent to the description by the Schrodinger equation. fJl
given by (VI.5.2), considered as a Hilbertian direct sum, is a complete Hilbert space; it is a
representation space of a unitary representation of SO(4) and of an irreducible unitary
representation of the groups SOC 4, 1) and SOC4, 2). In fJl the operator C 1 given by (VI.3.13)-
and therewith, by (VI.3.l7), the energy operator of the hydrogen atom-is self-adjoint. The
space of square-integrable solutions of the Schrodinger equation (3.1) with discrete eigenvalues
(the space of bound states) is not a complete Hilbert space. Remarkably, these distinct topo-
logical properties do not make any difference for the observable quantities.
The matrix element of the angular momentum operator Li = tijkQjP k in

the basis of generalized eigenvectors of the Qi is, according to (3.2), given by
1 0
<x ILil t/I) = -:-I tijkXr;--
uX
<x It/I). (3.5)
k
If the generalized position eigenvectors are labeled by the polar coordinates

(3.3), the differential operators on the right-hand side of (3.5) have to be
transformed from Cartesian to spherical coordinates; the results of that
somewhat lengthy transformation are
<xIL 1 It/1) = f(- sin 4> :e - cotecos4> o~)<re4>lt/I), (3.6 1 )
<xIL 2 It/1) = f(+ cos 4> :e - cot e sin 4> o~) <r e 4>lt/I), (3.6 2 )
1 0
<xIL 3 It/1) = i 04> <re4>lt/I)· (3.6 3)
From this one calculates
<x IUI t/I) = - (sin~ e 0~2 + Si~ e :e (sin e :e) ) <r e 4> It/I). (3.6)
For It/I) = In 1m) the equations (3.6 3 ) and (3.6) become

1 0
i 04> <re4>lnlm) = m<re4>lnlm) (3.6~)
and
1 -0 2 0 ( sm 0 )) <r e 4> In 1m) = l(l + 1) <r e 4> In 1m).

. e --
- ( -.-2- -2 + --.1-- --
sm e 04> sm e oe oe
(3.6')
Inserting (3.4) into (3.1) gives
11 0 ( 0 ) l(l + 1)
-2r20r r20r <re4> lnlm) + 2r2 <r84>lnlm)+V(r)<r84>lnlm)
=E<re4>lnlm). (3.7)
Using the expansion
Inlm) = fd3Xlx)<xlnlm)
= f r2 sin e dr de d4> Ire 4» <r e 4> In 1m) (3.8)
in the orthonormality condition

<n' l' m' I n 1m) = (jnn,(jll,(jmm' (3.9)
VII.3 Wave Functions of the Hydrogen Atom 237
gives the normalization condition for the transition coefficients:
f
r2 sin 8 dr d8 d4><n' l' m'l r 84» <r 8 4> In 1m) = <>nn'<>U'<>mm" (3.10)
Equation (3.7) for a given value of I is a differential equation in the variable

r only, while (3.6 3) and (3.6') are differential equations in the angular vari-
ables 8, 4> only. Therefore the transition coefficients are written
<r 8 4>ln 1m) = Rnz{r) llm(8, 4». (3.11 )
Equations (3.7), (3.6'), and (3.6 3) then lead to
(3.12)
and
(3.14)
For the hydrogen atom VA(r) = -air and E(n, l) = -a 2 /(2n 2 ). The normal-
ization condition (3.10) gives the following normalization conditions for
Rnl and lim:
(3.15)
and
The normalized solutions of (3.13) and (3.14) are the spherical harmonics
y. (8 A.) = (_1)1 21 + 1 (l + m)! im<J> . -m 8 (_d_)I-m sin 21 8 (3.17)

1m ,'V 21Z! 411: (l _ m)! e sm d cos 8 .
It should be noted that different authors use different phase conventions in

defining the llm(8, 4». Explicit expressions for the first few spherical
harmonics are given by llm(8,4» = 8 Im (8)e im 4>lfo where
8 00 (8) = j!, 8 20 (8) = A(3 cos 2 8 - 1),

8 10 (8) = A cos 8, 8 2 ±1(8) = +)11 sin 8 cos 8, (3.18)
8 1±1(8) = +j!sin 8, 8 2 ±2(8) = JH sin 2 8.
238 VII Alkali Atoms and the Schri:idinger Equation of One-Electron Atoms
The lIm( 8, ¢) form a complete set of single-valued functions over the unit
sphere:
Another useful relation involving the Y1m (8, ¢) is the coupling rule: 3
(21 1 + 1)(212 + 1)J1 /2

Y11m J8, ¢)YI2m2 (8, ¢) = ~ [ 4n(21 + 1)
x (llm112m211 m)(1 10 120110) flm(8, ¢). (3.20)
We also give here some useful relations which involve the spherical harmonics
along with the Legendre polynomials defined by
Pl~) = 21[!
1 (d)1
d~ (~
2
- 1)1 for I = 0, 1,2, ... , (3.21)
and the associated Legendre polynomials defined by
The spherical harmonics are related to the associated Legendre polynomials

by
(_1)m 21 + 1 (I - m)! eim<l>Pi(cos 8) for m 2 0

4n (I + m)! (3.23)
21 + 1 (I + m)! elm. PI-m(cos 8) lor
t'
- - m ~ 0,
4n (1- m)!
where for m ~ 0 we have made use of
(3.24)
When m = 0, (3.23) becomes
YlO (8, ¢) = !W-

--
I+ 1
4n
Plcos 8). (3.25)
Lastly, we state the addition theorem: Let nand n' be unit vectors with direc-
tions specified by (8, ¢) and (8', ¢'), respectively; then
3 See, e.g., Rose (1957), Section 14.

vn.3 Wave Functions of the Hydrogen Atom 239
The normalized solutions of Equation (3.12) for VA(r) = - ajr and

E(n, I) = - a2j(2n2) are
R () = (2a)3/2J-(~t=l)! (2ar)1 -ar/nL 21+ 1 (2ar) (3.27)

nlr n 2n[(n+/)!J3 n e n+1 n'
where L;~i l(p) are the associated Laguerre polynomials
(3.28)
a = 1jr Bohr is the inverse Bohr radius in units eVjc [cf. (2.19)J; in the usual
units in which rand r Bohr have the dimension cm, a is to be replaced by a'
of (2.20).
Figure 3.l(b) shows Rnlr) for some of the lower values of n and I. Figure
3.1(a) gives the probability distribution for finding the electron at r.
t·
0.5
. ..
',20 30
'
0.1
..,/).,.. ....
:.<- ...... '--.:.:
, ........
•• : :"0 .~:e.:
..•..•.4Q
..:..
5 20 25 30 rlrs
. IO
01 15
~
S slales (/ = 0)
~~
.. 0.2 ,----,"'- - , - - - - , . - - - , - - - , - - - ,
'"
.~
.;;;
<=:
~ 0.1 hl--+-*-+-:-=:-4---+---t--~
~
:E
«l
.n
o 10 15 20 25
~
01 5
-;; p states (/ = I)
~
~ 0.2 , - ---y---,----,.---,---,---,
O.I I--++-~~+-.,.....I----.~----i
o 5 10 15 20
d states (/ = 2) and 4/ state
(a)
Figure 3.1 (a) Charge distributions of the lowest states of the hydrogen atom. The
radial distance r is in units of the Bohr radius. The curves are labeled by the numbers nt.
,,
0.8 :
:10
0.7 ,
1
I
0.6 ~
.201 ,
0.5 I
1
32
30 32 r
12
------31
( b)
Figure 3.1 (b) The radial wave functions Rnt(r) for hydrogenic atoms for n = 1, 2, 3.
Each curve is labeled by nt. Note the effect of the centrifugal force in "pushing out"
the wave function from the nucleus. Note also that the functions have n - t - 1 nodes.
[Part (b) from Dicke & Wittke (1960), with permission.]
Knowledge of <x In 1m) allows an alternative computation of the matrix

element of V(Q) to the one given in the previous section; t.(n, I) =
<n 1m I V( Q)I n 1m) can now be calculated by inserting the complete general-
ized basis Ix):
t.(n,/) = Jd 3 X Jd 3 X'<nlm IX)<x IV(Q)IX') <x'lnlm)
= Jd 3 xl<nlmlx)1 2V(r).
Multiplying (3.28) by l'im(e, ¢), we obtain, according to (3.11), the transition

coefficient <r e¢ I n 1m). Substitution of <r e q) I n 1m) into the above
equation and use of (3.16) to perform the e, ¢ integrations then yield
t.(n, I) = Cna )3 ~:[~ 1+-I)I!j~ J r2 dr VCr) C:r)211 L~~jlC:r) 12 e-C:r).

(3.29)
These integrals can be evaluated for the various functions VCr) = -c/r 2 ,
VCr) = - c/r 3 , VCr) = - c/r 4 , ... and lead to the same results as given in
Equations (2.14)-(2.18). From Figure 3.1(a) we can make some qualitative
observations regarding the values t.(n, I). For a given value of n the probability
Problem 241
distribution increases and moves towards larger values of r as I increases.

Therefore, for a given value of n, - f(n, I = 0) is the largest, so that E(n, I = 0)
will be the lowest energy value, and with increasing I the E(n, l) will approach
the value - a 2 /(2n 2 ). That this is indeed the case can be seen from the ex-
perimental energy diagrams of the alkali atoms in the previous section.
Problem
1. Calculate the matrix elements of the position operator QK (K = -1,0, + 1) between
the SO(4) states In I m) of the hydrogen atom, i.e., calculate
<n l' m'IQKln 1m).

It is sufficient to do this for the third component Qo. (Hint: Derive the following relation:
Qi = -iH- 1 ,4; + iH- 1 [H,.t;(Q, P)],
and use the matrix elements for AJ These matrix elements are needed for the Stark
effect, Problem VIll.4.
CHAPTER VIII
Perturbation Theory
In this chapter perturbation theory is developed in a general form that is

easily adaptable to the discrete as well as the continuous spectrum. At the
end of Section VIII.1 the Wigner-Brillouin and the Rayleigh-Schrodinger
perturbation expansions are obtained as special cases. Section VIII.2 applies
this general procedure to the continuous spectrum and results in the Lipp-
mann-Schwinger equation.
VIII.I Perturbation of the Discrete Spectrum
This section will develop the general procedure called perturbation theory
for the evaluation of the eigenvalues E;. and eigenvectors 1 A),
HIA) = E;.IA), (Ll)
of an operator
H=K+V, (1.2)
which differs by a small perturbation V from an operator K whose spectrum
ta and eigenvectors a),
1
(1.3)
are known. K is called the "free" Hamiltonian and 1a) the fr~e eigenvector.
H is called the "exact" Hamiltonian and 1,.1,) the exact eigenvector. V is
called the perturbation Hamiltonian.
242
VIlLI Perturbation of the Discrete Spectrum 243
In this kind of perturbation-theoretical treatment it is assumed that to

every eigenvector la) of K there corresponds an eigenvector IA-(a» of H
such that E A(a) is given by fa plus a small correction term such that if V
is zero, then E A(a) = fa' In the case that fa is degenerate (i.e., that there are
other quantum numbers IJ besides a or A-), EA(a) will in general depend upon
these additional quantum numbers even though fa does not.
To have a concrete case in mind, one may assume that H, K, and V are
the operators of Section VII. 1 ; then a is the quantum number n which labels
the n2-fold degenerate eigenvalues of K: fa=n = -m e e 4 /(2h 2 n 2 ).1 Since
[Li' H] = 0 (due to [Li' V] = 0), one can choose a basis offree and a basis of
exact eigenvectors both to consist of eigenvectors ofL 2, L3: lal]) = lalm),
IA- 1]) = IA- I m). The problem then reduces to a multitude of non-degenerate
problems, one for each pair of values of the quantum numbers I] = (I m). If H
does not commute with L2, L3 and if one does not know of another set of
operators which forms c.s.c.o.s with Hand K separately, then the problem
cannot be reduced to non-degenerate problems and degenerate case pertur-
bation theory, which we do not discuss here, must be used. (cf. Problem 4).
Although at the moment we have in mind discrete values for fa and a,
we do not want to exclude the possibility that K has a continuous spectrum;2
in that case the summation La'
in the equations of this section, is to be re-
placed by an integration over the continuous spectrum (cf. Section VIII.2.)
We define the operator
G(E) == (EI - H)- 1 (J = unit operator), (1.4)
which depends upon the parameter E. G(E) is called the "operator for the
Green's function for the Hamiltonian" by the physicists, and is called the
"resolvant of the operator H" by the mathematicians. It is (intuitively) clear
that the task of finding the eigenvalues EA of H is equivalent to that of finding
the" singularities" of G(E).
We also define the operators
1
F(E) == G(E) geE) , (1.5)
and
R(E) == V F(E), (1.6)
where
geE) == L la)<aIGaCE),
a
(1.7)
GaCE) == <aIG(E)la). (1.8)

R(E) is called the level shift operator, for reasons that we now explain:
1 We revert to the usual cgs units in this chapter.

2 The following treatment of perturbation theory is based on the Goldberger-Watson
Collision Theory (1964), Section 8.1. It has been chosen because it displays the connection
between the cases of the discrete and the continuous spectrum.
244 VIII Perturbation Theory
Using the above definitions and the identity 3

(E - H)G(E) = I, (1.9)
one obtains
1 1
(E - H)F(E)la) = geE) la) = Ga(E) la) (1.10)
and, for the diagonal matrix elements of R(E),

Ra(E) = <al(H - K)F(E)la)
= <al(E - K)F(E)la) - <al(E - H)F(E)la)
1
= (E - fa)<aIF(E)la) - Ga(E)' (1.11)
It is easily shown that

<aIF(E)la) = 1, (1.12)
so that from (1.11) we have
1
E = fa + Ra(E) + Ga(E)· ( 1.13)
As E approaches an eigenvalue EA of H, Ga(E) becomes infinite and (1.10)

and (1.13) become
(1.14)
and
(1.15)
Though (1.15) is valid for any EA and fa' it is used in practice to determine
that value E; which is "closest" to fa' which we call EA(a). Thus Ra(E))
represents the level shift from the "unperturbed" energy value fa to the cor-
responding E A(a). The operator F(E) has been constructed such that F(E Ma)
is the operator that transforms la) into the corresponding l.-t(a» =
F(EA(a)la), which according to (1.14) is an eigenvector of H with the eigen-
value E A(a).
As already stated above, the assumption 4 that forms the basis of this
perturbation theory is that to every la'1) there corresponds a l.-t(a) '1)
3 E stands for E ·1; the unit operator in a product with a number will be omitted.
4 This assumption, which underlies perturbation theory, can be shown to be fulfilled for
certain V, as the following theorem states:
Let
with
M
IIK(k)fll ::::; k"=[(llfll + IIHfll) (k = 1,2, ... ),
r
VIII. I Perturbation of the Discrete Spectrum 245
(i.e., there are as many Ia 11) as there are IA(a) 11», but E ),(a) maybe a function
of 11 because Ra(E) does in general depend upon l1-e.g., in the example of
Section VII.1
(1.16)
will in general depend upon n, I, m. In the particular case [R(E), LJ = 0
it follows that Ra~(EA) does not depend upon m, but it will depend upon 1 in
addition to its dependence upon n. Thus all the la 11) belong to the energy
level fa of the unperturbed system, and under the influence of the perturbation
this level splits into sublevels E .l.(a)~ and the state Ia 11) of the unperturbed
system goes into the state IA(a) 11) of the perturbed system.
The level-shift Ra(E.l.(a», and thus E)'(a) , are evaluated by various sequences
of approximations. These various sequences are obtained in the following
way: For any two invertible operators A and B it is easily shown that
III 1
A = R + R(B - A)A' (1.17)
Choosing A = g(E)(E - H), B = g(E)(E - K - O(E», and hence

l/A = F(E), liB = (E - K - 0(E»-1 . llg(E), one obtains
1 1 1
F(E) = E _ K _ O(E) 9+ E _ K _ O(E) (V - O(E)F(E). (1.18)
Here O(E) may be any operator, but we shall only consider operators of
the form
O(E) = L Oa,(E)la')(a'l, (1.19)
a'
where Oa.(E) are numbers (i.e., O(E) is assumed to have the property
[O(E), KJ = 0).
Equation (1.18) is not defined on every vector, not even on every vector
Ia), for any value E; for example, the first term is only defined on those
Ia) and for those E for which
(E - fa - Oa(E»GaCE) i' O.
where the numbers M and r are positive constants. Then for IK I < r the eigenvalues E(K) of H
can be written
E(K) = E(O) + KE(1) + K2 E(2) + .. "
and the eigenvectors IE(K» of H can be written

IE(K» = IE(O) + KIE(t) + K2IE(2) + ....
For K --> 0 it follows that
H=K,
E(K = 0) = E(O) = f,,,
IE(K = 0» = IE(O) = la).

Applying (1.18) to la) and making use of (1.7) and (1.13) gives
I
F(E)la) = E _ K _ O(E) (E - fa - RaCE))la)
+E _ K 1_ O(E) (V - O(E))F(E)la). (1.20)
O(E) will be chosen such that this expression becomes the simplest possible
relation for F(E;.la))la); this can be done if F(E;.(a)) 1a) is written as la) plus
a correction term. Thus O(E) is to be chosen so that the first term becomes a). 1
Two choices of O(E) lead to particularly well-known perturbation expan-

sIons:
O(E) = Ra(E)la)<al, (1.21)
I.e.,
(l.21a)
and
O(E) = Ra(E)la)<al + (E - fa) L la')<a'l, (1.22)
a'*a
I.e.,
OAE) = RAE) for a' = a, Ou-eE) = E - fa for a' #- a. (1.22a)
In the case (1.21),
E - K - O(E)
and in the case (1.22),
E - K - O(E)
Thus in both cases (1.21) and (1.22) the first term on the right-hand side of
(1.20) becomes la). Using (1.6) and (1.12) it also follows that
<al(V - O(E))F(E)la) = o. (1.25)
Thus in both cases, (1.20) may be written as
F(E)la) = la) + a~a E _ fa' ~ OAE) la')<a'I(V - O(E))F(E)la), (1.26)
in which it is irrelevant whether the term with a' = a is included or excluded

in the sum. Inserting (1.21) into (1.26) gives
1
F(E)la) = la) + a~a E _ fa' Ic/)<a'i VF(E)la). (1.27)
VIII.) Perturbation of the Discrete Spectrum 247
This equation can be iterated by successive substitution into F(E) Ia) on

the right-hand side:
1
F(E)la) = la) + I ~-Ia')<a' I Via)
a'*a E - fa'
+ "L. - -1 --- 1- la')<a'lVla")<a"lVla) +. ". (1.28)

a' *a E - fa' E - fa"
a" *a
The level shift and therefore, by (1.15), the eigenvalue of H is obtained

from (1.28) by acting on the left with V, taking the scalar product with
Ia), and evaluating at E = E A(a):
1
Ei.(a) = fa + <alVla) + I <alVla')<a'lVla) + .... (1.29)
a'*a EA(a) - fa'
This provides an equation for E A(a) in terms of the unperturbed energy value
fa and the matrix elements of the perturbation V between the unperturbed
states la). Equation (1.29), derived from the choice (1.21) for O(E), con-
stitutes the Wigner-Brillouin form of the perturbation expansion.
The choice (1.22) of O(E) leads to the Rayleigh-Schrodinger perturbation
series. Inserting (1.22) into (1.26) and evaluating at E = E A(a) leads to
and for the level shift
(1.31)
Approximations of RaCE Ma) and F(E i.(a) Ia) to any desired order in the
perturbation V may be obtained by iterating (1.31) and (1.30). We denote by
R~ln) the nth-order approximation of RaCEA(a) and by F(n)[a) the nth-order
approximation of F(Ei.(a)la). To zeroth order,
(1.32)
To first order,
(1.33)
F(1)la) = la) + I 1 la')<a'I(V - R~9)F(O)la)

a' *a fa - fa'
1
= la) + I*a
a' fa - fa'
la')<a'i Via). (1.34)
To second order, one may verify that

1
R~2) = <alVla) + L
a',;,a Ea - Ea ,
<alVla')<a'lVla), (1.35)
The vectors F(n) Ia), or in general F(E Ma) Ia), are once again usually labeled
by the quantum numbers a. However, whereas for the unperturbed vectors
la) the labels are connected with eigenvalues of a complete set of operators
containing K of which these vectors are eigenvectors, the "exact" vectors
and the approximations F(1) Ia), ... obtained by the perturbation expansion
are in general not eigenvectors of the complete set of operators containing
H. Therefore we will often denote them by I(a», where those quantum
numbers which are bracketed do not refer to eigenvalues of operators
of which the I(a» are eigenvectors. The meaning of the quantum numbers
in the "exact" vectors is therefore only that these vectors are connected
with unperturbed eigenvectors whose eigenvalues are these quantum
numbers (or related to them).
VIII.2 Perturbation of the Continuous Spectrum-The

Lippmann-Schwinger EquationS
In the previous section it was nowhere stated that Ea is a discrete eigenvalue
and la) a proper eigenvector of K, though all equations were written as if
this were the case. The assumption of a discrete spectrum is in fact un-
necessary; we shall now discuss in more detail the case of the continuous
spectrum, which, with the tools of Section 11.7, is easily obtained by replacing
the proper vectors Ia) with the generalized vectors, and the discrete sums
by integrals.
For the following discussions, we have to make some assumptions regard-
ing the properties of the operators K and H = K + V. We shall assume that
spectrum K c spectrum H. (2.1)
This is a general enough assumption to incorporate the situations that occur
for physical problems. Usually for scattering and decay processes K has
only a continuous spectrum that is positive; the continuous spectrum of H
agrees with the continuous spectrum of K, and in addition H has a set of
negative discrete eigenvalues (and sometimes also positive discrete eigen-
values lying in the continuous spectrum) that are bounded from below.
5 The results of this section will not be used until Chapter XIV.
VIlI.2 Perturbation of the Continuous Spectrum 249
In this case of a continuous spectrum the level shift Ra is not an observable

quantity. However (1.15) still holds, for (as remarked in the previous section)
it is true for any value E). in the spectrum of H and any value fa in the spectrum
of K.
We shall consider (1.20) at the value E).(a) = fa' i.e., we shall calculate
F(E).(a»)la) for that value E).(a) E spectrum H which is identical to the eigen-
value fa E spectrum K corresponding to the generalized eigenvector Ia).
Such a value E).(a) exists in the spectrum of H because of the assumption (2.1).
We call this value EA(a) = Ea. For Ea, (1.15) becomes
(2.2)
If we again choose (1.21) for the operator O(E) we will again arrive at (1.27),
which in the continuous form is written
(2.3)
PJ means the principal-value integral, i.e., the integral over all a' with the
J
omission of an "infinitesimally small interval" around Ea, = Ea· da ... is
a symbolic way of expressing summation over the discrete label and inte-
gration with respect to a suitable weight function over the continuous label.
The label a in la) stands for the energy Ea and some additional labels 'la'
which may be discrete or continuous: la) = IEa'la). If the la) are "normal-
ized" so that
then
[cf. (1.4.7c ll ) and Chapter XIV]. Again (1.14) is fulfilled, i.e., F(Ea)la) is a
generalized eigenvector of H with eigenvalue Ea' This can also be seen directly
from (2.3): Multiplying (2.3) by (Ea - K) and using
Kia) = Eala)
and
J da' la')<a'i = I - la)<al,
one obtains
The right-hand side of this equation is according to (1.6) and (2.2) equai to
zero and one obtains:
(1.14')
We now make use of a well-known relation from the theory of generalized

functions, 6
~1~ = lim _1~ = p~ =+= in b(x). (2.5)
x ± iO ~>0 + x ± il'J x
Then (2.3) can be written
F(Ea)la) + Ida'
= la)
Ea - E a,
1
± I'0 la')<a'i VF(Ea)la)
± in Ida' b(Ea - Ea')la') <dIVF(Ea)la). (2.6)
Integrating and using (2.2), we see that the last term in (2.6) vanishes:
± inla)<aIVF(Ea)la) = ±inRaCEa)la) = O.
Because K Ia') = Ea,1 a'), Equation (2.6) may be written in the more usual
form
1
F(Ea)la) = la) + '0 V(F(Ea)la». (2.7)
Ea - K ±I
This is called the Lippmann-Schwinger equation. It is an equation for the
generalized eigen vectors
(2.8)
of the operator H with eigenvalue Ea; Ea is equal to the eigenvalue Ea of K
belonging to the generalized eigenvector Ia). Thus
Hlo:) = Hlo:(a» = Ealo:), (2.9)
The Lippmann-Schwinger equation is often written
Io:±) = Io:±(a» = la) + 1 '0 Vlo:±)· (2.7')

Ea - K ±I
Besides the solutions of the Lippmann-Schwinger equation 10: ±), there are
other basis vectors of interest. These are obtained from the form (2.3) of the
integral equation for F(Ea) Ia), which in operator form is written as
(2.10)
6 The proof of (2.5) may be found in Gelfand et al. (1964), Vol. 1, Chapter I, Section 2.4.
Further discussion of generalized functions of the type (x ± iO)" may be found in Chapter I,
Sections 3 and 4 of the same volume. Relations between generalized functions (distributions)
like (2.5) can be understood in the conventional sense when they are multiplied with a well-
behaved function cjJ(x) and integrated. Thus (2.5) means
f dX-~-:-
x ± 10
cjJ(x) = JdX ~x cjJ(x) + in fdX o(x)cjJ(x)
for all well-behaved functions cjJ(x). Analogously. relations between generalized eigenvectors
like (2.6) can be understood in the conventional sense when the scalar product is taken with a
well-beha ved vector cjJ E <l> (cf. Section 11.8).
Problems 251
where P indicates that the principle-value integral is to be taken. To dis-

tinguish the solution of this equation from the two solutions of the Lippmann-
Schwinger equation we denote it by IIX ) and write (2.1 0):
P
IIXP)=la)+ VIIXP). (2.10')
Ea - K
Problems
1. Let the energy operator be given by
H = K + IXQ3 + {JQ4,
where
(anharmonic oscillator).
(a) Calculate the eigenvalues to first order in the perturbation expansion.
(b) Calculate for {J = 0 the eigenvalues of H to the second order of the Rayleigh-
Schrodinger perturbation series. Compare the result with that of (a).
(c) Calculate for {J = 0 the eigenvalues of H to the second order of the Wigner-
Brillouin perturbation series. Compare the result with those of (a) and (b).
2. Let the exact Hamiltonian of a rotator be given by
H = 2i1 L + IXL3 + f3L

2
1
with [J ~ IX, f3 ~ hl2I. Let {II m): m = I, 1- 1, ... , - I} be a basis of eigenvectors of

U and L3 with eigenvalues h 2 l(l + 1) and hm, respectively:
(a) Using perturbation theory with K = (I/2I)U + IXL3 as the free Hamiltonian
and V = {JLl as the perturbation Hamiltonian, calculate the eigenvalues of
H up to the lowest order term involving {J.
(b) H may be expressed as
H = ~
21' L'2 + IXT3,
where the L; are angular momentum operators obtained by rotation from the
Li and where l' and IX' are constants. Express l' and IX' in terms of 1, IX, and f3
and determine the exact eigenvalues of H. Discuss whether the results obtained
in part (a) are consistent with the exact results.
3. Show that [Equation (1.14')]
HF(Ea)la) = EJ(EJla).
where according to Equation (2.3)
is the continuous-spectrum analogue to (1.27).

4. An external, static. homogeneous electric field tf) splits the energy levels of the
hydrogen atom (Stark effect). The perturbation Hamiltonian, obtained from corre-
spondence with the classical case, is
V= -D·tf)= -eQ·tf)
(tf) is measured in volts/cm.)

Using the result of Problem VII.I, calculate the energy eigenvalues to first order in
the perturbation V for the n = 1 and n = 2 levels of hydrogen. For n = 1, Equation (1.33)
may be used directly with la> = In = II = 0 m = 0>; however, because of the
degeneracy of the n = 2 level at zeroth order, one must diagonalize V within the n = 2
subspace in order to obtain basis vectors for which the perturbation expansion (e.g., the
second-order term) will converge. One will find that at first order the n = 2 level of the
H-atom splits under the influence of the electric field into three sublevels. Since this
splitting occurs at first order, it is called the linear Stark effect. Explain why there is
no linear Stark effect for the alkali (and other) atoms.
CHAPTER IX
Electron Spin
In Section IX.2 the doublet fine-structure splitting of one-electron atoms is

explained as a spin effect. To obtain quantitative results the interaction
Hamiltonian for the spin interaction is determined in Section IX.3. In Section
IX.3a the magnetic moment of the electron is determined by classical argu-
ments; in Section IX.3b the magnetic field, which acts on the electron
magnetic moment in an atom, is presented. In Section IX.4 these results
are used to calculate the fine-structure splitting. In Section IX.S selection
rules for dipole transitions are derived, and the chapter closes with some
general remarks concerning the visualization of quantum systems.
IX.1 Introduction
The existence of electron spin was suggested by the fine structure in atomic
spectra. (See the description of the experimental situation at the end of
Chapter VI.) The electron spin cannot be expressed in terms of the position
and momentum operators of the electron. If the electron is considered as a
physical object with translational and rotational degrees of freedom, then
spin is an observable that corresponds to the rotational degrees of freedom
in the same way that momentum corresponds to the translational degrees
of freedom. Thus the electron is an elementary rotator (Section V.I) with a
translational degree of freedom. In the preceding chapters we have ignored
the rotational degrees offreedom of the electron, as their contributions to the
energy ofthe electrons bound in atoms is small. We have shown in Chapter III
253
254 IX Electron Spin
Figure 1.1 Splitting of a particle beam after traversing a Stern-Gerlach magnet.
that angular momenta of any integral and half-integral value may exist and
that there is nothing peculiar about half-integral angular momenta. It turns
out that the spin (intrinsic angular momentum) of the electron is ~.
A direct verification of the electron-spin hypothesis came from the Stern -
Gerlach experiment. The experimental arrangement is shown in Figure 1.1 .
A strongly inhomogeneous magnetic field was produced between pole
pieces P1 and P 2 , of which P 2 had a sharp edge. Abeam of hydrogen atoms 1
in the ground state was sent closely past the sharp edge of pole piece P 2,
eventually hitting plate T. With no field the beam formed a narrow line on
the plate (Figure 1.1 dashed line). When the magnetic field was turned on
the line divided into two lines. (Figure 1.1 solid lines.)
Without going into the details for the description of this experiment (see
Chapter XIII), one can already see that the model of the hydrogen atom
described in Chapter VI cannot explain such a separation. According to that
model the ensemble of hydrogen atoms with lowest energy is in a pure state
W = A 91 (n= 1), (1.1)

because ;?,£(n = 1) is a one-dimensional space. According to the model of
Chapter VI, it would thus be impossible to separate the ensemble of the
original beam into two subensembles as has been done by the magnetic
field. The experimental separation into two subensembles shows that in-
stead of !?£(n = 1) we must have (at least) a two-dimensional space. If we
assume that the ensembles of the separated beams are in pure states, then
we must have an exactly two-dimensional space. Within a certain degree of
accuracy (including only the electron structure and disregarding possible
structure of the nucleus) this assumption has so far been upheld.
As the separation of the beam was caused by a magnetic field, one con-
jectures that the new observable, whose existence has been demonstrated
by the Stern-Gerlach experiment, must be connected with the magnetic
moment. As the magnetic moment is connected with rotating charges, one
would suspect that this new observable must be angular momentum.
1 The original experiment used silver atoms. The experiment has since been repeated with
other atoms. including hydrogen.
lX.2 The Fine Structure-Qualitative Considerations 255
IX.2 The Fine Structure-Qualitative Considerations
The angular momentum that causes the separation of the beam in the
Stern-Gerlach experiment cannot be the orbital angular momentum Li =
fijkQjP b because this is zero for the ground state of the hydrogen atom.
Furthermore, there are no two-dimensional spaces of states for the algebra
of orbital angular momentum since I = 0, 1,2, ... (Problem III.l). The two-
dimensional representation space of the algebra of spin angular momentum
is ~j~ 1/2 (Section 111.3), containing the two one-dimensional spaces of
states with opposite helicity. Denoting this two-dimensional space by 2 s , we
may write
2S = ~j~ 1/2 ~ ~j~ 1/2 r.:p
~j~ 1/2 (2.1)
~'(SO(2)S3) OJ] ~ -1/2 W J3 ~ + 1/2'
We thus conjecture that the electron is an elementary rotator and the

electron in the hydrogen atom is a combination of the physical systems of
an orbiting electron and this elementary rotator (spinning electron), the
space of states of which we call 2 Then the space of physical states of lowest
S•
energy of the hydrogen atom is
Yf(n = = = 1) ~(n 1)@,zs, (2.2)
and in general the nth space of the hydrogen atom is given by
Yf(n) = ~(n) @ ,zs. (2.3)
The orbital observables (functions of Qi' PJA i and L i , given by (VI.3.l),

(VI.3.3), and (VI.3.9), act in the space ~(n), and the new angular momenta,
which are called spin and denoted by Si' act in the space 2 S
•
To check whether our conjecture is in agreement with experiment, we

reduce the space Yf(n) with respect to total angular-momentum. According
to (VI.4.4), each space ~(n) is the direct sum ~(n) = Ll~6 EB [%1. Then
Yf(n) given by Equation (2.3) is the direct sum 2
Yf(n) ~
8(SO(3)L)
(nil EB ~/) @2s=nilEB (~/ @2S).(2.4)
1~ 0 1~ 0
~/ @ -is is the space of physical states of the combination of two elementary

rotators with angular momenta I and s = t. The orbital angular momenta
Li = fijkQjPk act in [%1, and the spin angular momenta Si act in -is. The total
angular momentum of the combined elementary rotators is then, according
to Equation (V.2.3),
Ji = L i @ I +I @Si' (2.5)
2 The symbol ~ means that the two spaces are equivalent under the action of
O'(SO(3),)
operators generated by the L i .
Using (V.2.32) we can reduce &£1 @ -t S to the sum of total-angular-momentum

eigenspaces
&£i= 1/2 if I = 0
{
&£1 @ -t
S
= &£i=l+ 1/2 EB &£i=I-I/2 otherwise. (2.6)
Whether the &£i or the &£1 @ -tS are the spaces of physical states depends, of
course, upon the preparability of these states. If physically preparable states
are eigenstates of the total angular momentum, then the &£i are the spaces
of physical states; if the physically preparable states are eigenstates of the
orbital and spin angular momentum, then the &£1 @ ZS are the spaces of
physical states. The experimental data show that the physical spaces are
the &£i, for each orbital-angular-momentum level (except the 1 = 0 level)
splits into two sublevels as required by (2.4) and (2.6).
From (2.4) and (2.6) it follows that
JIt'(n) = &£t/;.O) EB (&£t/;1) EB &£~/;l)) EB (&£~/;2) EB &£1/;2))
EB ... EB (&£n:~el) EB &£t;:;el))' (2.7)
For n = 1,
JIt'(n = 1) = &£t/;O), (2.8)
and there is one energy level corresponding to this two-dimensional space.
The two-dimensionality of JIt'(n = 1) allows us to explain the separation
of the beam in the Stern-Gerlach experiment into two subensembles, one
corresponding to the space ~l/;O)h= + 1/2 and the other to ~t/;O)iJ = - 1/2'
For n = 2,
(2.9)
and there are three energy levels, each of which corresponds to one of the
spaces in (2.9):
f7lN~o) ...... 2S 1 / 2 , f7lt/;l) ...... 2pl/2, and &£~/;1) ...... 2p 3/2. (2.10)
For n = 3,
JIt'(n = 3) = f7lt/;O) EB ~b/; 1) EB f7lW; 1) EB f7lW; 2) EB ~t,/; 2)' (2.11 )
which belong to the energy levels
3S 1 / 2, 3p1/2, 3p3f2, 3d 3 / 2, 3d 5/2,
respectively. For n = 4,5, ... the same arguments hold. These results reflect
precisely (if hyperfine splitting is neglected) the experimental data shown
in Figure V1.5.2 for hydrogen and Figures VII.2.l, VII.2.2, VI1.2.3 for
the alkali atoms. Thus to each energy level there corresponds an eigenspace
of the total angular momentum; eigenspaces of J2 are also eigenspaces of
H,and
[H, JJ = O. (2.12)
This equation, which we obtained here from the empirical facts, expresses the
rotational invariance of H and can theoretically be obtained from general
symmetry arguments.
We shall denote by )"forb the space f!ll given by (VI.S.2):
L
00
)"forb = E9 rJf(n). (2.13)

n~l
The operator H given by (VI.3.2) or (VI.3.S) is not the total-energy operator

when spin is taken into account; let us rename this approximate energy
operator H 0 :
p2 a
Ho = - - - (2.14)
2 Q
(or
p2 a
Ho = - - - + V(Q) (2.14')
2 Q
for the alkali atom). The space of physical states of the hydrogen atom in a
description that takes spin into account is then
)"f = )"forb Q9 its = C~l E9 rJf(n)) Q9 ZS = Jl E9 )"f(n). (2.15)
This description is the combination of the spinless hydrogen atom and an

elementary rotator. A basis system in this space is thus
(2.16)
where the
In / m) (n = 1,2,3, ... ;1 = 0, 1,2, ... ,n - l;m = -I, -/ + 1, ... , + I)
form a basis system for .?'f0rb, and the Is = 1S3) (S3 = -1, +1) from a basis
system for It' = f!A1/z. Each observable A in .Yf is written, according to the
basic postulate IV,
A -- "L. x Aspin
Ao rb to-.
(i) 'CY (i), (2.17)
i
where the A?nb act only in yt0rb and the A(l)in act only in t S [cf. (III.S.7)].
Orbital angular momentum and spin are special examples of this:
Li = L orb
i to-.
~ f
spin
= ( tijk Qorbporb)
j k to-. f spin =
~ Li to-.
~ f = (tijk Q j P k) to-.
~ f, (2.18)
(2.19)
where fS pin , forb are the unit operators in the spaces itS and )"forb, respectively.
(For typographical convenience we shall usually omit the superscripts
"orb" and" spin" if it is clear from the context in which space an operator
acts.) The total angular momentum of the combined system is the sum of
these two angular momenta,
(2.5)
The energy operator H can also be written in the form (2.17). Since H 0 @ I
already gives a very good approximation to the energy spectrum (cf. Figures
VI.5.1, VII.2.l, VII.2.2, and VII.2.3), we shall write the energy operator as
(Ho=Ho@I), (2.20)
where H 1 acts on the entire space :J'( = :J'(orb @ ,zS and is small in the sense
that its contributions to the energy values are small compared with the
values of H o. H 1 cannot be of the form H 1 @ I. If it were, no fine-structure
splitting between levels with the same I could occur, for H~rb could just be
added to Hgrb and the resultant H = (H 0 + H 1) @ I (for a given value of l)
would only lead to a shift of the energy levels. Neither can H 1 be of the form
I @ H 1, for ,zS is a very simple two-dimensional space: Every operator in
is can be written as a linear combination of the four operators I, S 1, S2, and
S3; thus H 1 would have to be H 1 = (lo I + If= 1 (liSi (ex i E q. The first
term would only lead to an overall shift of all energy values. The second
term (If=l (liS i), and consequently H 1 , does not commute with lk' because
eli' Sj] = iEijkSk (2.21)

as a consequence of (2.5) and of
[Si' Sj] = iEijkSk' (2.21a)

Thus it would follow that H 1 , and consequently H = Ho + H 1 , is not rota-
tionally invariant, in contradiction to (2.12). Hence the most general possible
form of H1 is
3
H1 = A @ I +I Bi @ Si' (2.22)
i= 1
where A and Bi are operators in :J'(0rb, i.e., functions of the operators Qi' Pi'
A is a scalar operator, while the Bi are the components of a vector operator
with respect to the L i . The former follows from the fact that every operator
in is = 3£ 1/ 2 can be written as a linear combination of the operators I, Sl,
S 2, and S 3' The latter follows from the requirement that
[H 1 , lJ = O.
which in turn follows from (2.12), (2.20), and [Ho, LJ = O.
To see this we calculate
3
I [B k @ Sk, lJ = [Bb L i] @ Sk + iEkijBk @ Sj'
k=l
which is zero only if
[Bk' LJ = iEkijBj' (2.22b)

i.e., if Bk is a vector operator with respect to the L i .
As
[Hb LJ ¥- 0, (2.22a)
which follows immediately from (2.22b), the basis vectors given by (2.16)
cannot be eigenvectors of H. Physically prep arable states appear to be
always energy eigenstates or mixtures of energy eigenstates (cf. Chapter
XII). Therefore we should use a basis of eigenvectors of the energy operator
H. Because of (2.12), eigenvectors of the total-angular-momentum op-
erators J2, J 3 can be eigenvectors of H. Therefore we use (V.2.12) to form
the new basis vectors
Inls = !jh> = L Inl13> ® Is = !S3><1I3S = !s3IJJ3>' (2.23)
13'3
where <113 !s3lj h>are the Clebsch-Gordan coefficients given in Table

V.2.1. The basis vectors (2.23) are eigenvectors of the c.s.c.o.
(2.24')
As 8 2 is!(! + 1) on the whole space Jr, we can ignore it and have in
H o , L2, J2, J 3 (2.24)
a c.s.c.o., of which the basis vectors of (2.23) are eigenvectors. The basis (2.23)
need not be a basis of eigenvectors of H. If the basis vectors (2.23) are not
eigenvectors of H, then eigenvectors of H can be obtained from (2.23) by
perturbation-theoretical calculations. As we shall show below
H, L2, J2, J3 (2.25)
is also a c.s.c.o., the eigenvectors of which we call
IEljj3>· (2.26)
The difference between (2.23) and (2.26) is that
Holn lj j3> = E~ln lj j3> (2.27)
where E~ = -a 2 /2n 2 , by (VI.5.3), and
HIE Ij j3> = EIE lj h> (2.28)
where E is a yet to be determined value. Thus the basis systems (2.23) and
(2.26) have all but one quantum number (the energy) in common and the
perturbation (H - H 0) will not affect the quantum numbers I, j, h. There-
fore after we have made the transformation (2.23) we can ignore the
additional quantum numbers of the state vectors and do perturbation
theory in one quantum number only, as described in Section VIII. I.
To prove that (2.25) is indeed a c.s.c.o. one needs only to show that
(2.29)
This cannot be derived from the general form (2.22), but constitutes a
condition on the A and B;. The justification of (2.29) comes from the con-
siderations in Section V.4 on parity, also expressed by (VI.5.9), and a new
assumption about H:
(2.30)
This assumption is called parity invariance (of H). Up in the spin space is
given by
U pis = ·b3) = 7rs ls = t S3), (2.31)

where l7rs l = 1. Equation (2.31) is (V.4.20) for the special case that j = s = t
and follows from (V.4.3) for the Sh i.e., from
(2.32)
Equations (V.4.20) and (V.4.21) withj = I for the orbital angular momentum
gIve
(2.33)
here'; may be any additional quantum number whose operator commutes
with Up, e.g., .; = nor'; = E. From (2.31) and (2.32) it follows that
Up(l'; 11 3 ) ® Is = tS3») = (-lYIJI'; 11 3 ) ® Is = tS3) (2.34)
(IJ = IJorb 7r s )· It then follows from (2.23) that
(2.35)
Consequently the eigenvalue of V is connected with the eigenvalue of Up,

and the quantum number I in the basis (2.23) may be considered the parity
quantum number. [The parity can have only the two values + 1 and -1;
thus for given n,j,h there exist two states of opposite parity. This has already
been expressed in (2.7) in the form that to every j there belong two l's, I =
j + t and I = j - t.J Thus if (2.30) holds, .; in (2.35) may be E, the eigenvalue
of H; from (2.30) it follows that E, I, j, j3 may label the states, and conse-
quently that (2.25) is a c.s.c.o., one which has now been shown to be identical
with the c.s.c.o.
(2.36)
Now that (2.29) has been established, it can be used to obtain some new
conditions on the operators A and Bi in (2.22). From (2.22) follows
[HI, VJ = 0 (2.37)
which does not give any new restrictions on A, but, along with (2.22b), does
restrict the Bi :
From this follows the restriction

(2.38)
where [f(Q, P), LJ = 0, i.e., f(Q, P) is a scalar operator with respect to
the L j • We will soon find by correspondence from the classical case that
f(Q, P) oc Q-3 for the spin-orbit contribution to HI [cf. (3.23)].
Although we have denoted the basis vectors (2.23) and (2.26) differently,
it may happen that they are identical (up to a phase). We have to distinguish
the two possibilities
[H 1, H oJ = 0, which implies [H, H oJ = 0, (2.39)
and
[H 1, HoJ i= 0, which implies [H, HoJ i= 0. (2.40)
If (2.39) holds, then the basis vectors (2.23) are also eigenvectors of H and
(2.41)
the eigenvalue of H may in general depend upon n, I, and j, and is given by
(2.42)
where (nlj is the eigenvalue of the operator H 1 in the state In 1j h):
Hdnljh) = Enliln1jj3)' (2.43)
If (2.40) holds, then (2.23) and (2.26) are different vectors, and (2.26) can be
determined from (2.23) by perturbation theory. The first-order perturba-
tion expression E~U for the eigenvalue of H is then given, according to
(VIII. 1.29) or (VIII. 1.33), by
Enli
(l) - EO
- n + Enli'
(1)
(2.44)
where E~U is the expectation value of H 1 in the states (2.23),
E~U = <n IjhI H llnljh)· (2.45)
IX.3 Fine-Structure Interaction

The calculation of quantitative results requires knowledge of H 1; H 1 can
be conjectured by some plausible physical arguments from the correspond-
ing classical situation. The fine-structure term in the energy operator con-
sists of two contributions: (1) a contribution caused by the interaction
between the magnetic moment of the electron and the magnetic field in
the electron rest frame due to the motion of the proton charge as seen in
this frame, and (2) a contribution caused by the variation of the mass of
the electron with its velocity. Both terms arise from relativistic kinematics;
although the last term is of the same order of magnitude as the magnetic-
moment term, it does not contribute to the splitting of the levels with the
same values of nand l. Besides the fine-structure terms there are other
contributions to the energy operator, which cause the Lamb shift and the
hyper fine structure, but we shall not consider those here.
IX.3a The Magnetic Moment of a Spinning

Particle in Classical Physics 3
The energy of a magnetic dipole of magnetic moment m in a magnetic field
B is
(3.1)
So we need to know the magnetic moment m of the spinning electron and
the magnetic field B in the electron's rest frame. In this subsection we shall
determine m, and in Section IX.3b we shall discuss B.
Before we consider the magnetic moment of a spinning particle, let us
recall the connection between the orbital angular momentum and the
magnetic moment caused by the revolution of a charged spinless particle
around a center. According to classical electromagnetic theory, the magnetic
dipole moment of an orbiting point charge at position x with mass me and
charge ( - e) moving with a velocity v is given by
morh = ~(_ e)x x v = _ _1_ el. (3.2)

2c 2mec
(Note that we consider negatively charged revolving particles like the
electron, and that e = + 4.803 X 10- 10 esu.) The connection between the
magnetic moment of a spinning charged particle and its internal angular
momentum (spin) s differs from (3.2) by a factor of gs ~ 2, the Lande factor,
i.e., the gyromagnetic ratio is e/(mec) and not e/(2m e c) as in (3.2). In order to
derive this factor we consider the spinning point particle.
A classical spinning particle is a physical system with two different
dynamical variables, the momentum p and the spin s. If x denotes the
position, then the total angular momentum j is given by
j = I +s= x x p + s. (3.3)
The intrinsic properties of this classical particle are assumed to be such that
it possesses a charge ( - e) and an intrinsic magnetic moment m. Then a given
external force F and external torque T can act upon it. The external force
may arise, for example, from an external electric or magnetic field (Lorentz
force); the external torque may arise, for example, from a magnetic field B
acting upon the intrinsic magnetic moment m:
T = m x B. (3.4)
The equations of motion are then
dp
dt = F, (3.5)
and
dj dl ds
-- = -- + -- = T (3.6)
dt dt dt tot'
3 Cor ben (1968).

where T tot is the total torque, i.e.,

T tot = x x F + T. (3.7)
Using (3.5)-(3.7) and the definition v = dx/dt, one obtains from (3.3):
ds
dt +v x p = T. (3.8)
If the momentum p and velocity v are parallel, or if either of them vanishes,

then
ds = T (3.9a)
dt
and one obtains by using (3.3) and (3.5) that
dl
dt = x x F. (3.9b)
This means that if the momentum vector p and the velocity vector v are
parallel, then the spin and orbital motions are independent; the spin motion
is determined by the torque T and the orbital motion is determined by the
force F. If, as is always assumed in nonrelativistic mechanics, p = mY, then
one can describe the orbital motion in terms of (3.9b) and the spin motion in
terms of (3.9a), and the spin and orbital motions become uncoupled. By
inserting (3.4) in (3.9a), one then obtains for the spin motion (of a particle
for which p and v are parallel)
ds
dt = m x B. (3.10)
Let us now assume that we have a particle that has no intrinsic magnetic
moment, i.e., m = 0; consequently, by (3.4), T = 0, where T is interpreted
as the torque acting on any intrinsic moment the particle may possess. But
let us assume that v x p i= 0,4 i.e., that we do not have the strictly non-
relativistic relation p = mv or the relativistic relation for free spinless
particles p = mv/(l - V2 /C 2 )1/2. Then (3.8) becomes
ds
-v x p, (3.11 )
dt
Suppose the particle is moving III a uniform constant magnetic field B.
The Lorentz force on the particle, which is equated with the time derivative
of the momentum, is then
dp -e
F = - = - v x B, (3.12)
dt c
Consequently
1d 2 dp e
2dt (p ) = P' dt = --P'V
c
x B. (3.13)
4 V X PI' 0 will occur if the time components SOl> S02, S03' which together with sij = EijkS k
make up the relativistic spin tensor s"" are not constant in time. This is always the case for an
observer for whom the particle is moving with a velocity v.
But, using (3.11) together with the supposition that B is a constant field
(dBldt = 0),
ed e ds e
--(s·B)=---·B= --v x p·B
edt edt c
e e
= - p x v· B = - p. v x B. (3.14)
c c
Adding (3.13) and (3.14), we conclude
e
tp2 +- s . B = const. (3.15)
c
If we suppose that the particle possesses a constant mass me and define
the quantity Jl as
e
Jl == - - s (3.16)
me c '
then we can write (3.15) as
p2
- - Jl' B = const. (3.17)
2me
This means that it is not the kinetic energy pZ 12me of the particle but rather
the left-hand side of (3.17) which is a constant of motion (unless two of the
vectors v, p, and B are parallel, in which case it would follow from (3.13)
that the kinetic energy is a constant of motion). The term - Jl' B therefore
represents an extra energy which the spinning particle, for which v is not
parallel to p, acquires in the magnetic field B. This extra term is written in the
usual form (3.1) for the energy of a magnetic moment Jl in the magnetic field
B. Thus, even though the particle (with pH v) has no intrinsic magnetic
moment upon which the magnetic field can act to give a torque, it "behaves"
such that there is a contribution to the energy in the magnetic field as if it
had a magnetic moment Jl given by (3.16) and a gyromagnetic ratio of elmec.
Equation (3.16) is similar to (3.2), but with an additional factor of 2 (the
value of the Lande factor g,,) on the right-hand side. We are thus led to the
conclusion that all charged spinning particles automatically possess a mag-
netic moment, given by (3.16), which is of relativistic origin. [The sign in
Equations (3.12)-(3.15) and (3.16) reverses for positively charged particles.]
If the electron is a quantum-mechanical particle with charge - e and
without intrinsic magnetic moment, then its magnetic-moment operator
should be given by the quantum-mechanical analogue of (3.16), i.e., by
e e
Ms= - - 5 = - g s - 5 (gs = 2), (3.18)
mec 2mec
where 5 is the spin operator. It turns out that (3.18) indeed gives a very
accurate description ofthe electron magnetic moment. But like every theoret-
ical description this is only an approximate description and there are devia-
tions. For electrons these deviations are small; for other particles, called
hadrons, the deviations from (3.16) are of the same order as (3.16) and are
interpreted as intrinsic magnetic moments due to an intrinsic structure. The
deviation from the 9.. = 2 value for the electron comes from the radiative
corrections of quantum electrodynamics and is of the same order as, and of
analogous origin to, the Lamb shift. Including these corrections up to the
second order gives
9s = 2( 1 + i; - O.328(;Y}
where Ct. = e2 /(hc) is the fine-structure constant; this value for 9. agrees with
experimental values up to the eighth decimal place. These correction terms
are small, due to the smallness of Ct., and the electrons are thus to a very good
approximation particles without an intrinsic structure.
The value 9. = 2 was first established as far back as 1915 by an experiment
of Einstein and de Haas, and was incorporated in the spin hypothesis put
forward around 1926. The existence of the radiative correction terms was
first discovered experimentally by Rabi and collaborators in 1947, and was
calculated within the framework of quantum electrodynamics by Schwinger
in 1948. The value 9. = 2 is also obtainable from the assumption of "minimal
coupling to the electromagnetic field," first used in the Dirac relativistic
wa ve equation for the electron and considered one of the great achievements
of the Dirac equation. The above argument shows that it is already a con-
sequence of classical considerations.
There are elementary particles with intrinsic structure; e.g., the proton
has an intrinsic magnetic moment
e e
Mp = 5.59 -2- S = -2- (2 + 3.59)S, (3.18p)
mpc mpc
i.e., a 9-value of gp = 5.59. The value of the magnetic moment in excess of

the value 2(e/(2m p c»S given by (3.16) is called the anomalous magnetic
moment; it is 1.7gej(m p c) for the proton. The neutron has no charge, and
therefore the magnetic moment given by (3.16) is also zero; however it has
an anomalous magnetic moment
(3.l8n)
IX.3b The Spin-Orbit Interaction Term
The determination of the magnitude of the magnetic field B must take into
account relativistic kinematical effects. We go into a coordinate system that
moves with the electron around the proton. In this system the electron is
at rest, and the proton charge moves with a velocity v that is equal in magni-
tude but opposite in direction to the electron velocity. This movement con-
stitutes a current. The magnetic field caused by a current of a single charge
+e moving with a velocity v is, according to the Biot-Savart law, given by
v x x
R(x) = +e-- (3.19)
3 cr
(e is measured in esu, B in gauss), where x is the vector of magnitude r from

the moving charge to the observation point. The angular momentum of the
electron is
1 = x x (- me v).
Consequently the magnetic field at the position of the electron caused by
the rotating proton is
(3.20)
Both (3.19) and (3.20) neglect relativistic effects. From (3.1) we thus obtain
(3.21)
for the energy of the magnetic moment in this field. The subscript "r.f.,"
denotes that we have used a rotating frame. If the frame rotates, there is an
extra contribution to the energy that reduces (3.21) by a factor of !. [This
factor is known as the "Thomas factor," and is caused by the "Thomas
precession"; a detailed calculation may be found in Jackson (1975), p. 364.]
Thus the energy of the moving spin magnetic moment in the magnetic field
of the proton is given by
E(m) __ _ e_ ~ I. (3.22)
1 - 2meC r 3 ms'
We obtain the quantum-mechanical expression corresponding to (3.22)

by the usual procedure of replacing the classical quantities 1 and ms by the
quantum-mechanical observables Land Ms given by (VI.3.1) and by (3.18):
(m) _ e2 1 . _ 1 e2 1 .
HI 2 2Q3 L S--222Q3LiSi
- +gS-4 (usualcgsumts)
meC me C
(h = 1). (3.23)
The quantity
eh (4.8 x 10- 10 esu)(6.6 x 10- 16 eV sec)
J.1 B = - - = -'-=--c:-::----:-::-;-;;,-----'--'-----:--:-::--,----:-::-~-:--'-
2mec 2(3.0 x 10 10 cm/sec)(9.1 x 10 28 g)
_8esueVsec2 058 cm l/2 sec
= 0.58 x 10 ---~- =. X 10- 8 eV 1/2
gcm g
= 0.58 x 10- 8 eV/gauss = 9.3 x 10- 21 erg/gauss
is called the Bohr magneton. The operator L· S of (3.23) is easily calculated
from (2.5),
J2 = (L + sf = L2 + S2 + 2L· S,
and the fact that
We then obtain for (3.23) that (in the usual cgs units)
1 e2
H\m) = ----rzQ-3 t(J 2 - L2 - iI). (3.24)
2mec
IX.3c The Kinematical Correction Term

We obtain the contribution arising from the relativistic mass effect by
expanding the relativistic expression for the kinetic energy
Po = J(m ec2)2 + C2p2
of the free electron in powers of p!(mec) [p = (p2)1/2]:
Po = meC2Jl + (~)2
meC
= me C2 (1 + ~ (~)2
2 meC
_~ (~)4 + ...)
8 meC
= mec2 + p2 _ ~ (p2
2me 2 2me
)2 meC
~ + ....
The rest energy me c2 of the electron is ignored because nonrelativistically

the energy is determined only up to an additive constant (the rest energy
would shift the energy levels by the same fixed amount). Thus classically
the kinetic energy with the first-order relativistic correction is
p2 1 (p 2 )2 1
Ekin = 2me -"2 2me mec2 '
in contrast to the usual expression
Going to quantum mechanics, we replace the numbers Pi by the operators

Pi and obtain
H~) = __1_ (~)2 = __1_ (Ho + e2)2 (3.25)

2mec2 2me 2mec2 Q
for the kinematical correction to
p2 e2
Ho=---· (2.14')
2me Q
[Both this equation for H 0 and (3.25) are in the usual cgs units, as opposed
to the units described above (VI.3.4). The total-energy operator with the
two corrections is then
H = Ho + Him) + H\k), (3.26)
where Him) and H\k) are given by (3.24) and (3.25), respectively.
IX.4 Fine Structure of Atomic Spectra
As HI = H\m) + H\k) contains the operators Qk (k = - 1, - 2, - 3), HI and

therefore H do not commute with H o. It therefore appears that we have the
situation described by (2.40), (2.44), (2.45), that In I j j3) are not the physical
eigenvectors and that E~l] given by (2.44) and (2.45) is only a first approxi-
mation.
In order to calculate the matrix elements (2.4S) we make use of (VIl.2.1S),
(VII.2.I6), (VII.2.17) with (VII.2.20). [Note that for I = 0 Equation (VII.2.I7)
is singular; however, for 1=0 the matrix element <111jhIL·Slnljh)
is zero~cf. (4.2) below~so that the matrix element ofL· S/Q3 is always well
defined.] We calculate the matrix element of the spin-orbit term H¥") and
find
1/2 for j = I + 1.
{
x -(1+1)/2 forj=l-t, (4.1)
because
t(J 2 - V - iI)ln Ij j3) = t(j(j + 1) - 1(1 + 1) -l)ln Ijj3>
1 I" {I for j = 1 + t,
= lin lh> x -(I + 1) forj = 1- t. (4.2)
According to (2.23),
<n I j h IQ- 31 n 1j j 3> = I I <.i h 1l/3 S = t s3) <ll~ s = t s31j h >
x <s3Is3><n 11 31Q- 31n ll~>.

Using the orthogonality relation (Y.2.IS) of the Clebsch-Gordan coefficients,
and the fact that the last matrix element is zero unless 13 = 13, which we
write as
we obtain
<nljj3IQ- 3Inljj3) = <nIIQ- 3Inl).
Thus from (4.1) and (VII.2.l7) together with (VII.2.20) it follows that
<n I"lh IH(m)1 I") _ 1 ( 1 )2 2m~e6 1

1 n lh - 2: me C e -1~n3(1 + 1)(1 + t)l
x -
1 {I for j = I + t,
(4.3)
2 -(I + 1) forj = 1 - t.
Recalling that, according to Equation (VI.S.12) with h = 1,
EO = _ mee~ ~_ = _ (e 2)2 m c2 _I_ (4.4)

n 1 2n 2 C e 2n 2 '
IXA Fine Structure of Atomic Spectra 269
(4.3) can be written as
EO (e 2)2 1 {1/(l + 1) for j = 1+ t

n C n(21+1) -1/1 forj=l-t.
(4.5)
The matrix element of the kinetic-energy term H\k) of (3.25) is
<n I j j3IH\k)ln lj h)
I
- - 22 {E~2 + 2E~e2<nIlQ-llnl) + e4 <nIlQ- 2Inl)}
me C
In this calculation (VII.2.15) and (VII.2.16) together with (VII.2.20) have

been used. Adding (4.5) and (4.6), we obtain for the fine-structure interaction
j: t)·
term
<nljhIHllnljj3) = - E~(e:r :2 (~- (4.7)
Therefore the matrix element of H = Ho + H 1 is (with h restored)
En1j = <n Ijj31Hln Ijj3) = E~(I + (~:r n12 C: -~)).

t (4.8)
The term in (4.8) that gives rise to the fine-structure splitting is the one
proportional to I/(j + t). Recalling that ry,,2 = (e 2/(hc»2 ~ (l/137?, we see
that this fine-structure term is four orders of magnitude smaller than E~
and gives the experimentally correct splitting between the energy levels of
different values for the quantum number j. This is shown in Figure VI.5.2,
where the superscripts stand for the total angular momentum value j. Thus
the energy levels are labeled by n(l)j where for (I) with I = 0, 1, 2, 3,4,5, ... one
uses the letters S, P, D. F, G, H, ... , respectively. For example, 2p 1/ 2 means
n = 2, 1= 1, j = t and 2p 3 / 2 means n = 2, I = 1, j = l Equation (4.8) is
independent of I and therefore cannot describe the splittings between the
levels 2S 1 / 2 and 2pl/2 or between the levels 3p3/2 and 3D 3 / 2 -i.e., between
states with the same value of j but different values of I-which are also
depicted in Figure VI.5.2. As mentioned before, this splitting, the Lamb
shift, has a different origin and is one order of magnitude smaller than the
splitting due to the spin.
Spin is not a fundamental notion of quantum theory. As discussed in
Chapter III, the angular momentum value of th naturally appears as a solu-
tion of the angular momentum commutation relations (lII.2.16), which in
turn are a consequence of rotational symmetry. The existence of angular

momentum with half-integer values can thus be derived from the funda-
mental postulates of quantum mechanics and the properties of symmetry
transformations (the Wigner theorem).
The spin value of y, is a property of the electron in particular, and not a
peculiarity of the general postulates of quantum mechanics. In the early
days of quantum theory, however, the effects of the spin were misconstrued
as being difficulties inherent in the general structure of quantum physics,
and this resulted in one of the most fascinating chapters in the history of
quantum mechanics.
That the angular momentum quantum number can have half-integer
values was first suggested by A. Lande (1921) in a systematics of the anoma-
lous Zeeman-effect (the splitting of the energy levels of atoms in an applied
magnetic field, see Problem 3). Thereafter (also in 1921) Heisenberg had the
idea that if an electron is added to an atom or ion, it imparts to the original
body of the atom a fraction ih of its angular momentum lh, and retains the
amount (/ - i)h. With this hypothesis he was able to explain the doublet
term splitting of the alkali atoms as it was later described by the above
formula (4.7).
The suggestion that this splitting was caused by a spin of the electron was
probably first made by R. Kronig (spring, 1925). He discussed it with Pauli
who rejected it as untenable. The main reason for refusing the spin hypoth-
esis was the missing factor of i (the Thomas factor) in (3.22) which led to a
discrepancy by a factor of 2 between the calculated doublet splitting (2 x the
right-hand side of Equation (4.7)) and the observed splitting (correctly
given by (4.7)). The fact that gs = 2 had by then already been well established.
Then in the fall of 1925, S. Goudsmit and G. E. Uhlenbeck, ignorant about
all this, published the first paper on the spin hypothesis and later, after
Heisenberg had drawn their attention to it, derived the doublet splitting with
the excess factor of 2. This convinced Bohr, but Pauli still objected to the
electron spin; he did not believe that there existed a mechanical explanation
of these effects. Finally, in 1926, L. H. Thomas showed that the additional
factor of i in the observed doublet splitting was a forgotten relativistic
effect. That a factor of i and not vic should be a relativistic effect came as a
great surprise even to the experts of relativity theory (Einstein included), but
even Pauli was at last convinced of the existence of the electron spin.
IX.5 Selection Rules

Selection rules for dipole tranSitIOns have been discussed previously, in
particular in Section V.4 for general angular momentum states and also
briefly in Section VI.5 for the hydrogen atom without spin. The selection
rules, i.e., the rules that tell when the matrix elements <n' I' j'j~ I Qi I n ijj.l >
are zero, follow from the property that Qi is a proper vector operator, i.e.,
IX.6 Remarks on the State of an Electron in Atoms 271
that it fulfills the relations

[L i , Qj] = itijkQb (5.1 a)
[1;, Qj] = itijkQb (5.1 b)
and
UpQiUP = -Qi· (5.2)
Equation (5.1 b) expresses the fact that Qj is not only a vector operator with
respect to the oribital angular momentum (an L-vector operator) but also a
vector operator with respect to total angular momentum (a J-vector op-
erator). Equation (5.1b) follows from J i = Li + Si together with (5.1a).
As a consequence of (5.1 b), the selection rules for dipole transitions between
the physical states In I j j3) for the hydrogen atom (and all other one-electron
atoms) are:
<n' I' j' j; IQiln lj h) = 0 unless j' = j + 1,j, or j - 1. (5.3)
From (5.2) it follows that
<n' I' j' j~ IQd n 1j h) = 0 unless n(l)n(l') = -1 (5.4)
[i.e., unless (_1/+ 1 = -1, as n(l) = (-1Y]. From (5.3) and the fact that
j' = I' ± ! and j = I ± !, it follows that the matrix element is zero unless
I' = 1 + 2, 1 + 1, I, 1 - 1, 1 - 2. Thus (5.4) can be written
<n'l'j'j~IQdnljj3)=O unless 1'=1±1. (5.5)
Equations (5.3), (5.4) or (5.5) give the selection rules for dipole transitions
in the hydrogen atom and all other one-electron atoms.
IX.6 Remarks on the State of an Electron in Atoms
We should close this part on the hydrogen atom with a remark. To conjecture
the algebraic structure that is the mathematical image of the hydrogen atom,
we made use of the classical particle picture of the Kepler problem, i.e.,
we had as the point of departure for our conjecture the picture of an electron
as a particle that moves in closed orbits around a center. And when we
discussed spin, we even implied some analogy of the spin with the rotation
of this particle around its own axis [in the comparison of (3.16) with (3.18),
for instance]. This was the classical picture, which is described by the mathe-
matical relations between the observables when they are not represented by
operators but by numbers. In quantum mechanics these mathematical
relations are operator relations, and are not the mathematical image of a
point particle revolving around a center and rotating around its own axis.
The electron in the hydrogen atom is not a particle, and the spin is not a
rotation around the particle's axis. A quantum-mechanical object was called
a particle when it was approximately a generalized eigenstate of the position
operator, because localization is the characteristic property of a particle.

The electron in the hydrogen atom is not in a generalized position eigenstate.
Equally wrong is the view (which originates from the misinterpretation of
the solution of the Schrodinger wave equation) that the electron in the
hydrogen atom is a standing plane wave, i.e., the classical picture com-
plementary to the particle picture. A quantum-mechanical object was called
a wave when it was a generalized eigenstate of the momentum operator
(Section 11.9), because this state has the characteristic property of a wave
(wave motion over all space). The electron in the hydrogen atom is not in a
generalized momentum eigenstate. Thus neither of the complementary clas-
sical pictures of the electron is applicable to the electron in the hydrogen
atom. In the hydrogen atom the electron does not appear as a wave or
as a particle, but in a form different from both, namely, as an angular-
momentum and energy eigenstate, in which it has neither a definite position
(particle) nor a definite momentum (wave), but does have a definite angular
momentum (rotator). The classical picture that comes closest to this is that
of a standing spherical wave.
Problems
1. (a) Show that every operator on is is a linear combination of the operators
I, Sj' Sz, and S3; as stated in the paragraph preceding Equation (2.21).
(b) Check to see whether there are other solutions of
[the equation immediately preceding (2.38)J besides (2.38).
2. The proton possess a magnetic moment
(a) Determine the Hamiltonian that describes the interaction of the electron's
spin with the proton's spin in the hydrogen atom (hyperfine interaction).
(b) Treating the spin-spin interaction Hamiltonian in part (a) as a further perturba-
tion term on the hydrogen-atom Hamiltonian, determine the matrix elements
of this spin-spin term between states In I j j 3) of I = 0 and I = 1. See how this
correction compares with the fine-structure correction.
3. An external magnetic field breaks the rotational symmetry so that the Hamiltonian
H for an atom in the external field no longer commutes with the angular momentum
operators J i .
(a) How does such an external symmetry breaking field affect the energy spectrum?
The Hamiltonian for the hydrogen-atom or the Na-atom in a uniform external
magnetic field B is given by
H = Ho + H + H'
j
Problems 273
whereH o + HI is the rotationally invariant Hamiltonian forB = O:Ho = (l/2m)p2+ V

and HI gives the fine structure. H' describes the effect of B and is given by
H' = IL 0 8'(L + 2S)
where 110 = eh/2mc (Bohr magneton).
(b) Calculate the spectrum of H for the case that the magnetic field is very weak
so that H' can be treated as a small perturbation of H 0 + HI and HI'
(c) List the dipole transitions between all energy levels of the 3p2 band and the
3S 2 band of the Na-atom in a weak magnetic field. (Zeeman effect)
CHAPTER X
Indistinguishable Particles
In this chapter the basic assumption about the combination of identical

physical systems is conjectured from the indistinguishability of these systems.
X.I Introduction
The quantum-mechanical systems that we have considered so far have

consisted of only one constituent of the same kind. The hydrogen atom was
considered as one electron in an electric field. The vibrating diatomic
molecule was described as the problem of one oscillator. The rotating
diatomic molecule was reduced to the problem of one system rotating around
a center. When the vibrating and rotating diatomic molecule was con-
sidered, it was described as the combination of one rotator and one oscil-
lator, i.e., as one rotating oscillator. These systems are called one-particle
systems. Thus one-particle systems are systems that consist of only one
constituent of the" same kind."
We now want to consider many-particle systems, i.e., systems that are the
combination of many (N = 2, 3, ... ) one-particle systems of the same kind.
Let £1' £2' ... , £N be the spaces of physical states of the first, second, ... ,
Nth one-particle system. (All of the spaces £i are identical; the subscript
serves only to identify a space with a particular particle.) We would expect
from the basic postulate IVa that the N-particle system, which is the
combination of these N one-particle systems, will have as its space of states
the direct product space
f> = £1 @ £2 @ ... ® £N· (Ll)
274
X.I Introduction 275
The algebra of the N-particle system would be the direct product of the
algebras of observables of the one-particle systems, i.e., the algebra is the
set of all operators
(1.2)
where Ai is an element of the algebra of observables in Yt;. (For example,
if these N one-particle systems do not interact with each other, then all
elements of the algebra of observables on i) are of the kind A = Al ® I ®
I ® ... ® I + I ® A2 ® I ® ... ® 1+ .,. + I ® I ® ... ® I ® AN')
That the N-particle system would be described by (1.1) we would expect
from classical considerations. Classical particles can, in principle, be num-
bered; i.e., we can imagine that each particle is given at some instant a label,
and we can then follow the subsequent motion of each particle in its path
and identify a given particle at any subsequent instant. This is not possible,
however, for quantum-mechanical systems. Following the motion of each
particle of a quantum-mechanical system means performing a series of
position measurements. Each measurement-according to the axiom 111-
changes the state of the quantum-mechanical system in an uncontrollable
fashion: If we have localized the particle in the neighborhood of a certain
point, then we do not know what its momentum is and we do not know
where it will go. The concept of path does not exist for a quantum-mechanical
system (uncertainty principle). Thus in quantum mechanics it is, even in
principle, impossible to follow each of a number of identical particles and
thereby distinguish them. (By identical particles we mean particles that have
the same observable values.) Thus: Identical quantum-mechanical particles
are indistinguishable.
We shall now give a precise, mathematical formulation of indistinguish-
ability and then formulate the consequences as another basic postulate
of quantum mechanics. Let I~)i denote a basis in the space :Ifi, i.e., the one
symbol ~i stands for the full set of quantum numbers (eigenvalues of a c.s.c.o.)
necessary to label the basis system of :Ifi. [For the sake of definiteness, one
may assume that the N particles are electrons in a Coulomb field; then each
:Ifi is the space.Yt of (lX.2.15), and I~)i = InJijiji3>'] The basis in i) of
(1.1) is then given by
1~1 ~2'" ~N> = 1~1>1 ® 1~2>2 ® ... ® I~N>N' (1.3)

Suppose we are given N objects (elements) in a certain order,
(~l> ~2"'" ~N);
such an arrangement is called a permutation. These N elements can be
written in a different order, (rJI,rJZ, ... ,rJN); this is called a permutation of
the N objects. There are N! different permutations of N objects. One par-
ticular permutation can be considered to be the" original" or "natural" or
"standard" one. All other permutations can be obtained from this original
one by changing the order in which the objects appear. It is clear that the
operation of changing the order is specified by the resulting permutation;
276 X Indistingui~hable Particles
therefore this operation is also called a permutation. Thus we may consider

the permutation operation (or just "permutation") P that changes
(~b~2' ... '~N) into (11b112,···, 11N). For example, we can consider the
permutation P 12 that changes (~I' ~2'···' ~N) into (111 = ~2' 112 = ~I'
113 = ~3' ... , 11N = ~N)· Or we can consider the permutation Pij that changes
(~I'···' ~i'···' ~j, ••• , ~N) into (~I'···' ~j' ••• ' ~i'···' ~N) by interchanging
the ith and jth elements; such permutations, which consist of the exchange
of the position of two elements, are called transpositions. Every permutation
can be obtained by a finite number of transpositions; e.g., the permutation
(~2' ~3' ~I' ~4'···' ~N) may be obtained by the transposition P l3 followed
by the transposition P 12 : (~l' ~2' ~3' ~4'···) ~ (~3' ~2' ~b ~4'···) ~
(~2' ~3' ~b ~4'·· .). While the decomposition of a permutation into suc-
cessive transpositions is not unique, the number of transpositions will always
be even or odd, depending upon the particular permutation. A permutation
is odd (with respect to the original permutation) if it is obtained from the
original permutation by an odd number of transpositions; it is even if it
is obtained by an even number of transpositions.
Now let the N objects be the N sets of quantum numbers (~I, ~2' ... ) =
(nl II jl jl3, n212 j2 h3' ... ). Each permutation (111, 112, ... , 11N), or each
permutation operation P:(~b~2' ... '~N)f---+(11l,112, ... ,11N) can then be
represented in the space f> by a linear operator IP defined by
(1.4)
For example, the transposition P 12 is represented by the operator IP 12 with
lPuI~l ~2··· ~N) = 1~2 ~l ... ~N)· (1.5)
The operators IP will be chosen to be unitary. 1
[The set {P} of all permutations of N objects forms a group known as

the symmetric (or permutation) group, while the set {IP} of all rep-
resenting operators forms a representation of the permutation
group. If all operators IP are unitary, then the representation is
called a unitary representation of the permutation group. Because
the permutation group has a finite number of elements (is finite
and consequently compact), every representation of the permuta-
tion group, according to a theorem, may be considered unitary.
We shall make use of only one property of the permutation group,
to be stated below, and shall not require any group theory.]
Let A10 denote the projection operator onto the one-dimension subspace
spanned by 10 = I ~ I ~2 ... ~N), and let A1q > denote the projector onto the
one-dimensional subspace spanned by 111) = 1111112·· . 11N). Because of (1.4),
we have the connection
(1.6)
1 The reason for the requirement that IP' be unitary follows from the fact that IP' is a sym-
metry transformation. A brief justification of this will be given in Appendix to Section XIX.2.
Now, due to indistinguishability, A 10 and AI~) can represent a (pure) physical

state only if
(1.7)
or, using (1.6),
(1.8)
Thus of the one-dimensional subspaces of the direct product i) of (Ll),
only those for which (1.8) is fulfilled can represent physical states. From this
we conclude that, because of indistinguishability, it is not the whole direct-
product space i) of (Ll) but only a subspace of it that is the space of physical
states. Also, it is not the whole direct-product algebra of operators given by
(1.2) but only a subalgebra of it that is the algebra of observables. We now
want to determine the physical subspace of (Ll). It is clear that, in general,
the direct-product basis (1.3) is not a suitable basis [(1.8) means IP 1'1) ex:. 1'1),
which cannot be fulfilled if all the quantum numbers are different]; and that,
in general, the physical states are represented by linear combinations of (1.3).
Let tf; be a vector of the physical subspace [in general, a linear combination
of (1.3)], and let IP be a permutation operator. Then, if the particles are
indistinguishable, tf; and X = 1Ptf; [or A", and Ax = Ap ",] represent the same
physical state. The expectation value of every observable A must therefore
be the same for tf; and for X, i.e.,
(1.9)
for every observable A. As tf; is an arbitrary vector of the physical subspace,
we conclude that
(1.10)
for every observable A. Hence for any A E.91 and for any permutation
operator IP,
[IP, A] = O. (1.11)
[Equation (1.11) follows immediately from (LlO) for IP unitary,

which, as mentioned above, can be assumed for the permutation
group. For IP non unitary, one must take linear combinations
tf;! + iatf;2, tf;! + atf;2 in order to deduce (1.11).]
Equation (1.11) is the mathematical formulation of the statement that

identical particles are indistinguishable. From this mathematical formula-
tion of indistinguishability, one can deduce that the vectors of the physical
subspace of i) must fulfill either
1Ptf; = + tf; for all IP (1.12)
or
1Ptf; = (- 1ytf; where {pp IS~s even. if IP. is even,

odd If IP IS odd.
(1.13)
278 X Indistinguishable Particles
The l/J fulfilling (1.12) are called symmetric, while those fulfilling (1.13) are
called antisymmetric.
To deduce this, we need the mathematical formulation of a physically
obvious property of the algebra .91 = {A} of observables A. This can be
formulated in the following way:z .91 = {A} contains a complete set of
commuting operators. To justify this condition physically, we recall that we
called a state" pure" (up to a certain accuracy) if there is no observable
whose measurement allowed the separation of an ensemble in this state into
two or more subensembles. If such a separation is possible (as in the case of
the hydrogen atoms in the ground state), it leads to the introduction of a new
quantum number and therewith to the introduction of a new observable
(in the case of the hydrogen atom, spin). Consequently one must enlarge the
algebra of observables to accommodate this new observable. What is a pure
state with respect to a certain accuracy need not be a pure state with respect
to a higher accuracy. But up to every desired (and observable) accuracy,
every pure state is completely specified by a set of quantum numbers that
are connected-with observables. (In fact, the algebra of observables is con-
jectured from these quantum numbers and observables.) Each label of a
vector is connected with an observable, and the observables whose eigen-
values label the vectors form a complete set. [As we have seen, the statement
that a certain set of commuting operators is a complete system is a physical
statement, conjectured from the physical properties of the system, and not a
mathematical statement (cf. Chapter IV)]. Therefore the apparent occurrence
of pure states not fulfilling (1.12) or (1.13) is always an indication that the
available set of quantum numbers is not complete and that there exist other
quantum numbers completing this set that have not yet been uncovered.
Let us denote by AI> A z , ... , An a complete system of commuting opera-
tors of the algebra of observables, by Ia) = Ia 1 a 2 ••• an) the corresponding
eigenvectors, and by Ala) the projectors onto the subspaces spanned by the
la). Ala) is the observable whose expectation value gives the probability
of obtaining a 1 ,a 2 , ••• ,a" in a measurement of A 1 ,A z , ... ,An- As ob-
servables, the Ala) must commute with all permutations IP in accord with
(1.11):
(1.14)
From (1.14) it follows that Ia) must be an eigenvector of alllP.
PROOF. Equation (1.14) applied to a vector lift) gives IP' Ala) Iift) = Ala)lP'lift)·
Using Ala) = la)<al, we see that IP'la)<alift) = la)<allP'lift), and hence that
IP'la) = la) <allP'lift). D (1.15)

<alift)
The possible eigenvalues, <a lIP Il/J)/<a Il/J), follow from the following prop-
erty (which we shall not derive here) of the permutation group: The permuta-
2 It is this condition that is relaxed when parastatistics are allowed.

tion group has two one-dimensional representations: (1) the symmetrical

representation in which all permutations are represented by the unit oper-
ator 1, i.e.,
[plllf;) = + I If; ) for all permutations [pl; (1.16)
and (2) the antisymmetrical representation in which all even permutations

are represented by the operator 1 and all odd permutations are represented
by the operator -1, i.e.,
[plllf;) = (-lYllf;) where {p ~s even.if [pl.is even, (1.17)

P IS odd If [pl IS odd.
We recall that (Section 111.3) an irreducible-representation space (ladder

representation) of the algebra of W} is a representation space that is obtained
by applying all [pl to one element of the space. The fact that Ia) is an eigen-
vector of all [pl means that Ia) spans a one-dimensional irreducible-representa-
tion space. Therefore, according to the above property of the permutation
group, we must have either
[plla) = + la) for all [pl (1.18)

or
[pll a) = ( -1)P Ia) for all [pl. (1.19)
Since this is true for every basis vector Ia), and as every vector IIf; ) of the
representation space of the algebra of observables (i.e., of the physical sub-
space of ~) can be written as a linear combination of the basis vectors, we
conclude that either (1.12) or (1.13) must be fulfilled.
Let us denote by x~ that subspace of ~ which consists of the symmetric
vectors,
x~ = {11f;)E~: [plllf;) = 11f;)}' (1.20)
and let us denote by xf!... that subspace of ~ which consists of the anti-
symmetric vectors,
xf!... = {11f;)E~: [plllf;) = (-lYllf;)}· (1.21)
We can then formulate the statement, which we have derived from indis-
tinguishability (1.11), in the following way: The space of physical states of
N identical quantum-mechanical systems is either the antisymmetric space
X~ or the symmetric space x~.
A basis of nonnormalized vectors in x~ is given by
10+ = 1~1 ~2··· ~N)+ = L [pl1~1 ~2··· ~N)' (1.22)

IP'
where I ~ 1 ~2 . . . ~N) is given by (1.3) and LIP' is the sum over all permutations
[pl of the N objects ~ 1, ~ 2, ... , ~N. It is clear that the order of the quantum
280 X Indistinguishable Particles
numbers ~ 1> ~2' ... , ~N in 10+ is irrelevant. To prove that the 10+ are
symmetric, we calculate for an arbitrary permutation IP I that
IPIIO+ = L IPIIPI~I ~2 ... ~N> = L lP'I~1 ~2 ... ~N> = 10+,
!l' !l"
where we have set 1P' = IPIIP and where the sum over alllP becomes a sum
over alllP'. The latter follows from the fact that if IP I is fixed and IP runs over
all permutations, then 1P' runs over all permutations. Thus for any permuta-
tion IP 1,
1P110+ = 10+· (1.23)
A basis of nonnormalized vectors in Yf~ is given by
10- = I~I ~2 ... ~N>- = L (-l)PIPI~1 ~2 ... ~N>' (1.24)
!l'
To prove that they are antisymmetric, we first note that ( -1)p' = ( -l)P+ PI =
( -1)P( -1)PI and hence that ( -1)P = ( -1)PI( -l)P'. Then
1P110- = L (-l)PIPIIPI~1 ~2'" ~N>
!l'
= L (-l)PI( -1)P'1P'1~1 ~2'" ~N> = (-l)PIIO- (1.25)

!l"
for any permutation IP I' If the 1~;)i are normalized vectors in Yf;, then the
1~1 ~2 ... ~N> are normalized in f), i.e., <~1 ~2 ... ~NI~1 ~2 ... ~N> = I. Con-
sequently the 10+ and the 10- are not normalized; the normalizing factors
are calculated in Problem 1.
We state the result of the preceding considerations as a new basic postulate
(axiom) of quantum mechanics:
IVb. The space of physical states of N identical quantum-mechanical

systems (particles) is £~ if their angular momentum (spin) has an integral
value, and is £r:.. if their angular momentum has a half-integral value.
Particles whose space of physical states is Yf~ are called bosons, and particles
having Yf~ as their space of physical states are called fermions. Bosons are
said to obey Bose statistics, while fermions are said to obey Fermi statistics.
The above axiom IVb then states that half-integral-spin particles are
fermions, that integral-spin particles are bosons, and that there are no other
particles obeying some other "parastatistics" (which would have to belong to
higher-dimensional representations of the permutation group). This axiom
has been confirmed in all cases where it has been investigated (electrons,
protons, neutrons are fermions; pions, photons, phonons, alpha particles
are bosons). We remark that the axiom IVb has been described, to a large
extent, from indistinguishability (1.11), which in turn was deduced from
previously formulated basic postulates of quantum mechanics. The part
that has ,not been deduced is the connection between spin and statistics.
The Pauli principle in its original form follows immediately from IVb.
The Pauli exclusion principle states: The quantum numbers of two or more
electrons can never entirely agree.
Problem 281
Problem
1. Calculate the length of the vectors
1~1 ~2··· ~N>+ = L: !P1~1> ® 1~2> ® ... ® I~N>

II'
and
1~1 ~2 ... ~N>- = L:(-1)P!P1~1> ® 1~2> ® ... ® I~N>'
II'
where the I~) are normalized basis vectors in the space .Yf and the sum runs over all
permutations !P. The factor (-I)P is + 1 or -1 according as !P is an even or odd per-
mutation.
CHAPTER XI
Two-Electron Systems-
The Helium Atom
The system with two electrons is studied in this chapter. Section XU shows
that the space of physical states of the helium atom is the sum of the para-
helium and the orthohelium spaces. In Section XI.2 the ionization thresholds
(i.e., the energy values Enoo at which one electron is in the nth level and the
other is just dissociating from the atom) are determined, and the energy
levels below the first ionization threshold are discussed. Section XI.4 discusses
the energy levels above the first ionization threshold without considering
the interaction between these levels and the energy continuum of the (He + , e)
system.
XI.1 The Two Antisymmetric Subspaces of the Helium Atom
We shall illustrate the consequences of the basic assumption IVb with the
example of two electrons in a Coulomb field. This is the simplest nontrivial
case; although it does not demonstrate the full extent of IVb, it is mathe-
matically simple and does not require the introduction of further properties
of the representations of the permutation group.
The energy of two classical spinless particles of mass me with charge
-e that move in the field of a central charge Ze is given by
1 Ze 2 Ze 2 e2
E = - (pi + pD - - - - + -, (Ll)
2me r1 r2 r 12
where r 1 and r 2 are the distances of the first and second charges - e from
282
XLI The Two Antisymmetric Subspaces of the Helium Atom 283
the charge Ze, and where r 12 is the distance between these two charges.
This system is the classical analogue of the helium atom (or of any two-
electron ion if Z #- 2). We obtain the energy operator of the helium atom
by the usual procedure of replacing the numbers P.i' X. i , r. = (X;)1/2, and
r 12 = «Xl - X2)2)1/2 by the operators p.i> Q,j, Q. = (Q;)1/2, and Q12 =
«Ql - Q2)2)1/2, respectively. In addition we have to add a term Hl that
describes the influence of the electron spin. Thus we have
1
(Pi + PD - -Ze - -Ze + -e + H 1
2 2 2
H =- (1.2)
2me Ql Q2 Q12
for the energy operator of the helium atom (Z = 2), which we write in the
form
(l.3a)
H 0 is the Hamiltonian operator corresponding to the classical Hamiltonian
given by (1.1):
(l.3b)
where
( l.3c)
p2 az
h=~-- (l.3d)
• 2 Q.
and
(l.3e)
The operator H and all the other operators act in the space
~ = £1 (8)£2' (1.4)
where £. is the space of the system that consists of one electron in the
Coulomb field of the charge Ze.
For two objects ~l and ~2 there are only 2! = 2 permutations, (~l' ~z)
and (~2' ~l); therefore for a fixed set of quantum numbers ~l and ~2 with
~l #- ~2 there are only two basis vectors in ~: 1~1~2> and 1~2 ~l>. The
normalized symmetric and antisymmetric vectors for this fixed set of quan-
tum numbers are, according to (X.1.22) and (X.1.24),
1
I~l ~2>+ = J2 (I~l ~2> + 1~2 ~l» (1.5)
and
(1.6)
284 XI Two-Electron Systems-The Helium Atom
Thus for a fixed set of quantum numbers ~ 1 and ~ 2 with ~ 1 i= ~ 2 we have a

two-dimensional space spanned alternatively by I~ 1 ~2)' I~2 ~ 1) or by
1~1 ~2)+' 1~1 ~2)-' If ~1 and ~2 are fixed with ~1 = ~2' we have a one-
dimensional space spanned by I~ 1 ~ 2) = I~ 1 ~ 2) + . The space i) is spanned
by 1~1 ~2)' where ~1 and ~2 independently can take any of the possible sets
of values (nljj3) (n=I,2, ... ;j=0,1, ... ,n-l; j3=-j, -j+l, ... ,
j; I = j ± 1).1 The space £': is spanned by all the vectors I ~ 1 ~ 2> _, and the
space £'~ is spanned by all the vectors I~ 1~ 2> + . Consequently
(1.7)
i.e., the product space is the direct sum of the symmetric and antisymmetric
subspaces.
[Equation (1.7) is a particular feature of the case N = 2; for N> 2,

i) of (X. 1.1) is not the direct sum of the symmetric and antisymmetric
subspaces (X. 1.20) and (X. 1.21). Rather
i) = £'~ EEl £'~ EEl £'~, EEl £'~2 EEl ....
where there are a finite number of terms, as many as the ways in

which one can write N as a sum Ii Ni of positive integers N i • For
N = 2 there are two ways possible (2 and 1 + 1), and consequently
there are only two different terms, as given by (1.7).]
According to the axiom IVb, or the Pauli principle, only the subspace
£': of (1.7) is the space of physical states for the two-electron system. (Were
we considering a two-boson system, £'~ would be the space of physical
states.) To construct £': and to find the properties of the algebra of ob-
servables in £':, we proceed in the following way: Each £'a (IX = 1,2) is,
according to (lX.2.l5), written as
(1.8)
where £,~rb is the space in which the orbital observables (i.e., the observables
that are obtained as functions of the P ai and QaJ of the IXth electron act, and
where -t~ is the space in which the spin observables of the IXth electron act.
We now combine the orbital and spin spaces of the two electrons separately,
i.e., we form
(1.9)
and
(LlO)
We then find the symmetric and antisymmetric subspaces of £,orb 2 and -ts2
separately by the same procedure as described above; the ~a in (1.5) and
We shall now neglect HI and undertake a detailed construction of all

four spaces on the right-hand side of (1.15). We start with the spin spaces
{~ and {s~, as these are much simpler than the orbital spaces. The space
{S2 = {~ ® {~ is the direct product of 2 two-dimensional spaces and is thus
four-dimensional. Its direct-product basis is given by the four vectors
(1.16)
One easily finds the antisymmetric and symmetric combinations of these

four vectors. The symmetric ones are
li)l ® li)2' l-i)l ® l-i)2'
and ~ (li)l ® l-i)2 + l-t)l ® It)2)' (1.17)
while the single antisymmetric vector is
~ (li)l ® l-i)2 - l-i)l ® li)2)· (1.18)
The four vectors of (1.17) and (1.18) are orthonormal and therefore con-
stitute a basis in {s2; consequently the three symmetric vectors span the sym-
mectric space {s:,
which is therefore three-dimensional, and the vector (1.18)
spans the antisymmetric space iS~, which is therefore one-dimensional.
{s2 = ~SI = 1/2 ® ~S2 = 1/2 is the space of the combination of two elementary
rotators. We can thus apply the results of Section V.2 and define the operator
of total spin,
Si = Sli ® I +I ® S2i (i=1,2,3). (1.19)
Equation (V.2.32) then tells us that
(1.20)
i.e., the total spin is 8 = 1 or 8 = 0, and we can introduce in is2 the basis
188 3 ) with
S2188 3 ) = 8(8 + 1) 188 3 ),} 8 = 0,8 3 = 0;
{ (1.21)
S3188 3) = 83 188 3) 8 = 1,8 3 = -1,0, 1.
The observables Si commute with the permutation operators (in this case
the transposition operator iP' 12)' since
iP'12 Si = iP'12(Sli ® I + I ® S2J = (I ® S2i + Sli ® I)iP'12 = SiiP'12·
(1.22)
The two subspaces is: and i~ are eigenspaces of iP'12 corresponding to the
eigenvalues + 1 and -1, respectively. Because of (1.22), Si cannot transform
out of either is: or i~ , i.e., Si leaves i~ and i~ invariant. Thus Si leaves in-
variant [fts = 1 and [fts = 0 on the one hand, and leaves invariant is: and i~
on the other hand. Also, [fts= 1 is three-dimensional, as is i~, and ~s=o is
XLI The Two Antisymmetric Subspaces of the Helium Atom 285
(1.6) stand for ¢~rb = (na la la3) when considering ~orb2 and for ¢~ = sa3
when considering 'is2. In this way we arrive at
(1.11)
and
(1.12)
The total space 5 is then given by2
5= ~orb~ ® 'is"
= (~~b2 ® 'is:) ffi (Jt'~b2 ® 'is~) ffi (Jt'~b2 ® 'is:) EB (Jt'~b2 (8) 'is~). (1.13)
In Problem 1 it is shown that the symmetric subspace is

Jt'~ = (Jt'~b2 (8) 'is:) ffi (Jt'~b2 ® 'i~) (1.14)
while the antisymmetric subspace is shown to be
~:= (Jt'~b2 (8) 'i~) ffi (Jt'~b2 ® 'i~). (1.15)
Thus the space of physical states Jt': is the direct sum of two spaces: one is
the space of symmetric orbital states and antisymmetric spin states, and the
other is the space of antisymmetric orbital states and symmetric spin states.
As has already been discussed (Section IX.2) in the case of the hydrogen
atom, the basis vectors that are eigenvectors of the (total) spin and of the
(total) orbital angular momentum are not a physical basis, because it is the
(total) angular momentum that is the physical observable, and not the spin
or orbital angular momentum. Thus to obtain the physical states one has
to form those linear combinations of the direct product states I~rb ~~rb> + ®
Is 13 S23 >_ (in Jt'~b2 (8) 'is~) and I~~rb ~~rb> _ ® Is 13 S23 > + (in Jt'~b2 ® 'is:)
that are eigenstates of the total angular momentum. Furthermore, it will
turn out that the spaces ~,,:b2 (8) 'is~ and ~~~v (8) 'is: are not eigenspaces of
the energy operator H of (1.3a); the reason IS that HI does not commute with
the operator of total spin Si = Sli + S2i> but 'is: and 'i~ are eigenspaces of
S2. If the physical states are eigenstates of H (as all experimental data con-
firm) then the physical state vectors are elements of neither Jt''':b 2 ® 'is: nor
~~b2 (8) 'is:, but are linear combinations with a small component in one of
the spaces and a large component in the other. Thus the reduction of Jt':
into the direct sum given by (1.15) is only approximately physical; the
subspaces ~~b2 ® 'i~ and Jt'~b2 ® 'is: are spaces of physical states only to
the extent that the contribution of HI (the spin-orbit interaction) to H can
be neglected. As in the case of the hydrogen atom, this will turn out to be a
very good approximation.
2 We wish to stress again that the appearance of only symmetric or anti symmetric subspaces
is a particular feature of N = 2; for N > 2, higher-dimensional representations (J, of the per-
mutation group in £",b N and in ~.,N also have to be considered. The antisymmetric space £~
then contains not only the spaces £"fb N ® 't,±N as in (1.15), but also contains all direct-product
spaces of the form £,:',b N ® it,where (J' is the irreducible representation of the permutation
group that is "associated" with the irreducible representation (J in such a way that £~'bN ® i~':
is an antisymmetric subspace of fl.
XI.2 Discrete Energy Levels of Helium 287
one-dimensional, as is .z~; consequently ~F 1 = .zs: and ~s;o = .z~. We

can therefore write (1.15) as
(1.23)
The space of physical states (neglecting the spin-orbit interaction) is thus
the direct sum of a space in which the total spin is zero (space of singlet
states) and a space in which the total spin is one (space of triplet states).
This, as we shall discuss below, is the explanation for para- and orthohelium,
first given by Heisenberg in 1926.
XI.2 Discrete Energy Levels of Helium
We will now investigate the structure of the orbital spaces in (1.11). In each
£~rb (0: = 1,2) we have a reducible representation of the algebra 6'(SO(4))
of the orbital angular momentum
(2.1)
and the Lenz vector
A oi = (- 2h a) -1/2 (1 2(ikl
{P 01' L}
ok +T
az Qoi )
' (2.2)
which, according to (lX.2.13), is given by
L EB ~,(n).
00
£~rb = (2.3)
n; 1
In the direct-product space

£orb 2 = £~rb ® £~rb . (2.4)
we have a representation of the algebra of orbital angular momentum and
the Lenz vector given by
Li = Lli ® I + I Q9 L 2i ,
(2.5)
Ai = Ali ® I + I ® A 2i .
The operators of (2.5) are defined in analogy to the definition (V.2.3) of the
total angular momentum ofthe combined system of two elementary rotators.
It is easy to see that Li and Ai obey the same commutation relations as Lai
and A ai , i.e., the commutation relations of 6'(SO(4)):
We can define the operator

C 1 == C ll ® J + I ® C w (2.7)
where, similarly to (V 1.3.1 3) and (VI.3.l7),
(2.8)
We can further define the operator

C 2 == AiLi (2.9)
in analogy to (VI.3.14) for the hydrogen atom. The operators C I and C 2
commute with Ai and L i . However, these operators do not fulfill the same
relations as the operators Cal and C a2 . In particular, C I is not related to
the energy operator H 0 by as simple a relation as (VI.3.17), and C 2 is no
longer zero. To see the latter, insert (2.5) into (2.9); one then calculates that
C 2 = LliAli ® I + I ® L2iA2i + Ali ® L2i + Lli ® A 2i ,
which gives [making use of (VI.3.18)]
(2.10)
This is not, in general, identically zero.
The energy operator of the helium atom in the approximation in which
the influence of the spin is neglected is given by (l.3b):
Ho = Hoo + w. (2.11 )
It is customary to consider first the term (1.3c):
1
Hoo = - (hI ® I + I ® h2)· (2.12)
me
H 00 is the energy operator for a system of two non interacting electrons in
the (nuclear) Coulomb field. As it is very unrealistic to neglect the Coulomb
interaction between the two electrons (which is of the same "strength" as
the Coulomb interaction between each electron and the nucleus), H 00 is a
very poor approximation to the energy operator of the helium atom. Thus
we cannot expect that the spectrum of H 00 will give a good approximation
of the energy spectrum of the helium atom. As we will see later, it happens
that the qualitative features of H 00 agree with those of the energy operator
H 0 = H 00 + W; this justifies the usual treatment of first considering
H 00 separately.
Using (2.8) and (2.12), H 00 can be written
Hoo = - -a~- ( 1 ® I +I ® 1 ). (2.13)

2me C ll +I C21 +I
The spectrum of Hoo is easily found. We introduce into yt'orb 2 = yt'~rb ®
yt'~rb the direct product basis
(2.14)
As, according to (1.23), we want to know the spectrum of H 00 in the sym-
metric subspace yt''':b 2 and in the anti symmetric subspace yt'~b2, we in-
troduce in yt'orb 2 = yt',,:b 2 EB yt'~b2 the basis system of symmetric and anti-
symmetric vectors
(2.15 ±)
Equation (2.15+) gives the basis system in yto;b, while (2.15-) gives the
basis system in yt~b . The vectors (2.15 ±) are eigenvectors of H 00 and
together form a complete basis system in yt0rb 2. The spectrum of Hoo is
therefore obtained by applying (2.13) to (2.15 ±); the result is
H 00)
spectrum ( - 00,
- = Enn " ( - I2 +-2
= - RHe I )' (2.16)
me n n'
where
4
R" = ai:2 = 4(m e e ) = 4R" = 544 eV
He 2me 2h2 .. (2.17)
R~e differs from the Rydberg constant for the hydrogen atom by a factor of 4.
It is the Rydberg constant for a one-electron system in the Coulomb field
of a charge Ze = 2e, i.e., for the He + ion. Its value in cm - I (wave-number
units or inverse wavelength units) is
1 54.4 eV
RHe = 2nhc R~e = 12.40 x 10 5 eV cm = 4.39 x 10 5 cm- I .
The basis vectors (2.15 ±) are not eigenvectors of the total orbital angular
momentum L 2 and L 3 , and are not eigenvectors of the energy operator H 0
(recall that H 0 is the energy operator if the contribution of the spin H I is
neglected). If the physical states are eigenstates of the energy operator H,
they are very closely eigenstates of H o' Eigenstates of H 0 can be eigenstates
of the total orbital angular momentum L2 and L 3 , but can be neither the
direct-product states (2.14) of angular momentum I and l' nor the particular
linear combinations (2.15 ±) of those direct product states. [The states
(2.15 ±) are eigenstates of Li + q. But
[Li + LL Qi2] = -2[L lk L 2b (Qli - Q2i)(Qli - Q2J]
= - 4if kii L lkQ2jQli - QliQ 2j L 2k ) i= O.
Consequently the vectors (2.15 ±) are eigenvectors of an operator that does
not commute with H o .] One therefore has to couple the angular momenta I
and l' in (2.15 ±), according to the rules of Section V.2, to obtain eigenvectors
of the total orbital angular momentum. These eigenvectors, which are formed
as linear combinations of (2.15 ±), are not yet H 0 eigenstates but still H 00
eigenstates; H 0 eigenstates can then be formed as linear combinations with
the same value of total orbital angular momentum.
Instead of approximating the helium atom by a model system that con-
sists of two non interacting electrons that move in the Coulomb field of a
doubly charged nucleus, one can try to approximate it by a model system
that consists of one electron moving in the electric field that is formed by the
doubly charged nucleus and the other electron. This model is certainly much
more realistic if the one electron is-in the classical picture-far away from
the nucleus and from the other electron, which are close together. To obtain
this approximation we write
I h 1+ H el
H 0=- 2=-1 h2+ H el1> (2.18)
me me
where
(2.19)
More precisely,
and
Equation (2.18) is exact, but H~I is not an operator in £,~rb; it would be an

operator in £,~rb if we made the replacement
(2.20)
In our classical picture this would mean that the electron that is close to the
nucleus is really at the position of the nucleus. With the asymptotic re-
placement (2.20), the energy operator for the electron far away from the
nucleus has the asymptotic form
p2
H:1 = -2" + -
(2e-Q + -Q
2 e 2 ) p2 e2 1
~ -2" - -Q = - H"HYdr' (2.21)
me ,,12 me "me
i.e., the energy operator for the distant electron is the same as the energy
operator of the electron in the hydrogen atom. In the approximation (2.21)
the energy operator H 0 is approximated by
(2.22)
where H"Hydr is the hydrogen-atom Hamiltonian corresponding to (VI.3.5).

Letting n' ~ n, the spectrum of R 0 is then
spect rum H- 0 = E-nn' = - R"He n21 -

R" n1/2
= -R,,(i + _1)
n2 n,2
00 _
~ Enn , - -R
,,(4 + 4)
n2 n /2 . (2.23)
A more realistic approximation for Ho of the form (2.18) is to make the

replacement
(2.24)
where Z~1 is a number between one and two that expresses the screening of
the nuclear Coulomb field by the nearby electron. A possible approximation

for H 0 is thus
H 0 -+ H O(Zcfrl = H~'(Zefrl + H~'(Zefrl (2.25)
The value of Z~~ should be different for different states of the helium atom.
For the ground state one would expect Z~~ = Z~~; for the state with one
electron in the ground state and the other in a very high state one would
expect, according to (2.22), Z~~ ~ 2 and Z~~ ~ 1. Thus for different sub-
spaces we have different operators H O(Zefrl'
Let us now consider the subspace of states of the helium atom that can be
characterized-in the classical picture-in the following way: One of the
electrons has just dissociated itself from the helium atom, i.e., the system
consists of a He + ion together with an electron of zero relative energy. Let
us call the subspace of these states Yf'00; on this subspace H 0 can be ap-
proximated very well by
(2.26)
where H:' is approximated by (2.21). The subspace Yf'oo is then defined as

that subspace of Yf'0rb 2 on which
H:' = O. (2.27)
Let us denote the eigenvectors of H:' in Yf'~rb by
(2.28)
in distinction to the eigenvectors In 113 >a of h,. These eigenvectors have the
property [under the approximation (2.21)]
(2.29)
and are eigenstates of L; a.nd L,3' The eigenvectors on which (2.27) is

fulfilled are denoted by
100 113}" (2.30)
The direct-product basis in Yf'00 is therefore given by
(2.31)
We introduce in Yf'oo the basis of symmetric and antisymmetric vectors
(2.32 ±)
On Yf'00 the energy operator (2.26) has, according to (2.27), the following
spectrum:
(2.33)
For Hoof (2.26) applied to (2.32 ±) gives
h In 11 )2
1M ( 100 l' 1~}1 ® ~ + H~11 00 l' 1~}1 ® In 113 )2
3
V2 me '--v---'
=0
± !!J.. In 113\ ® 100 l' I~h ± In 11 3)1 ® H~11 00 1'l~}2)
me '--v---'
=0
a2 1
= - 2m:n2 j2 (100 l' 1~}1 ® In 11 3)2 ± In 11 3)1 ® 100 l' I~h)·
The Enoo are the energy values of the helium atom when one of the electrons
has zero energy and are also the energy values at which one of the electrons
dissociates from the atom (ionization thresholds). We expect that these are
the highest energy values for a given quantum number n, i.e., for states in
which one of the electrons has the principal quantum number n, because ifthe
other electron has the principle quantum number n', its relative energy with
the He + ion will be negative. En 00 is called the nth ionization threshold; it is the
value at which one of the electrons is just dissociating itself from the atom
while the other is in the nth energy level.
Comparing (2.33) with the spectrum of Hoo given by (2.16), we observe
that
E2n~ -+ Enoo from below as n' ~ 00. (2.34)
Comparing it with (2.23), we observe
Enn , ~ Enoo from below as n' ~ 00. (2.35)
From our above considerations we expect the eigenvalue Enn , of H 0 to lie
between the eigenvalues of H 00 and i1 0:
(2.36)
In the energy diagram of helium we first draw the energy levels Enoo
(n = 1,2, 3, ... ). The energy levels E nn , lie, according to (2.34), (2.35), and
(2.36), below En 00 , and lie closer together as n' becomes higher.
The larger the values of n', the closer are the eigenvalues of Hoo to the
energy values of the helium atom. Consequently for large n' and a given n,
the space spanned by the vectors (2.15+) and (2.15-) represents to a good
approximation the space of physical states. In the classical picture of the
helium atom, increasing values of n' mean that the "second" electron is
further away from the nucleus and from the "first" electron, so that the in-
fluence of the interaction term e2/r 12 is small; thus the classical picture supports
the result described above. For decreasing values of n' the interaction term
e2 1r 12 becomes more important, and we expect larger deviations ofthe energy
value of the helium atom from the value E~~. We expect the largest deviation
for the states in which both electrons have the principal quantum number
n = n' = 1. In this case the interaction term is expected to be largest (both
electrons are closest to the nucleus and to each other), and E?? is a very poor
approximation to the eigenvalue of H o. In this case, therefore, the space

spanned by (2.15 +) with n = n' = 1 is far from being the space of (orbital)
physical states. Thus for low values of nand n' the helium atom cannot be
considered as consisting of two independent electrons (independent-particle
approximation) in a Coulomb field. The physical states are not states in
which one electron is in one particular state and the other electron is in
another particular state; rather it is only the helium atom considered as a
whole that is in the physical states.
We now consider the space of states (which we call A1 ) of the helium atom
that have energy values below the value E 100' The space of eigenstates of H 0
with eigenvalue E 100 describes the state of the helium atom in which one
electron is in the lowest state (n' = 1, I' = 0, I~ = 0) and the other electron
is just dissociating from the helium atom. The eigenvectors of H 00 in this
subspace A1 can be obtained from the independent-particle states (2.15 ±):
"'~U3 = fi (In' = 11' = 0 I~ = 0>1 ® In 113>2

+ In 113>1 ® In' = 11' = °=
I~ 0>2)' (2.37)
Y~U3 = fi(ln' = II' = °I~ = 0>1 ® I n113>2

-ln1l3>1 ® In' = 11' = 0 I~ = 0>2)' (2.38)
The "'~II' span the space of symmetric states of Ap which we call A1 +; and
the Y~1I3 span the space of antisymmetric states of A1, which we call A1_·
That is, "'~U3 and Y~U3 are eigenstates of Hoo with eigenvalue E~~= 1:
HOO"'~1I3 = -R~e(1 + ~2)"'~1I3' (2.39)
(2.40)
As I' = 0 and I~ = 0, "'~1I3 and Y~1I3 are also eigenvectors of L2 and L3 with
eigenvalues 1(1 + 1) and 13 , respectively. Let us denote by "'n1l3 and Ynll3 the
eigenvectors of H o , L2, and L3 in A1 + and in A1 - , respectively. The eigen-
vectors "'n1l3 and Ynll3 of Ho can be obtained from the eigenvectors "'~113 and
Y~1I3 of Hoo by acting on them with a (unitary) operator that does not change
the eigenvalues of L 2, L 3 , and Ifl> 12 :
(2.41)
Ynll3 = U- (nl)Y~U3' (2.42)
This transformation U±(nl) may be different for different values of n and I;
U±(nl) does not depend upon 13 , as [H o , La = O. Equations (2.41) and
(2.42) show that the eigenvectors of H 0 can still be characterized by the
principal quantum number n, though they might be quite different from the
eigenvectors of H 00 with eigenvalues E~~ = 1 = - R~e(1 + l/n 2 ). For large
Figure 2.1 Splitting of the eigenvalues of Hoo under the influence of W.
values ofn, U±(nl) is very close to the unit operator. The eigenvalues of Ho
in ~1 we denote by E:1 (I) and En1 (1); in general they will depend upon I:
(2.43)
H 0 Yn1l3 = E;;I (l)Yn1l3' (2.44)
We are not that much interested in the calculation of the exact values of E;I (I),
as we want only to obtain a qualitative understanding of the energy spectrum
of the helium atom. Therefore it is not necessary to know the operators
U ± (n/). (They are connected with the interaction term Wand can be calculated
by approximation methods.)
The important conclusion for the qualitative understanding of the energy
spectrum that we draw from (2.41) and (2.42) is that to each n2 -fold de-
generate eigenvalue E~f of HO~ on ~ 1 + there correspond the n energy values
E:1 (I), the interaction term W slpits E~f into n sublevels as shown in Figure 2.1.
And to each n2 -fold degenerate eigenvalue E~f of HO~ on ~I_ there correspond
the n energy values E;;I (/). As the result of our above consideration we obtain
the following energy spectrum ofthe helium atom below E 100 (Figures 2.2 and
2.3): The lowest energy value is Eil (I = 0) in the symmetric subspace Yf,,:b\
.j( 0:h2 ® :!Il.\::O 0 .'/{ 0:h2 ® .11.\ = 1
.. parahelium" .. orthohelium"
first ionization threshold E 100
Or-~~--"--~~--T-------~~-------
Is3s Is3p Is3d
Is2s Is2p ----
-5 Is2s
-10
-15
-20
-25
IS Ip ID 3S 3p 3D
Figure 2.2 Energy levels of the helium atom below the lowest ionization threshold.
Singlet states Triplet states

.. para helium ...it°"b' "orthohclium" .),(ocb'
eV eV
IS Ip ID IF 3S 3p 3D 3F
o
°
-3.40
--
-6.05 -
-
-
-10 - -10
-13.6 1
- =
- - =
-- - - -
-20
= - -20
- -
-30 -30
-40 -40
-50 -50
-54.44
- - - - -- -- -- --
-
- -
-60
- -60
-70 -70
E:I(I = 0)
-79.0
-80 80
Figure 2.3 Term diagram for energy levels of helium. (The part below the first ionization
threshold is again given in Figure 2.2 which is the term diagram of the helium that one
usually finds.) The energy levels above the first ionization threshold E 100 lie in the
continuous spectrum.
which is connected with En

= - Ri'.e(1 + 1). There is no corresponding
energy value in the antisymmetric subspace £~b2, because YUh = O. The
values of Ell and E7~ as calculated from (2.23) and (2.16) are
- 00
Ell = -68.05 eV and Ell = -108.8 eV.
The experimentally measured value of Eil(i = 0), the double ionization
energy of helium (minimum energy required to free both electrons), is:
Eil(i = 0) = -79.0eV,
which shows that H 00 is really a very poor approximation to the energy
operator H o' The energy difference
E loo - Et1(0) = -54.4 eV + 79.0 eV = 24.6 eV
is the energy that is necessary to dissociate one electron from the helium
atom when it is in the ground state; it is called the "ionization energy" or
"ionization potential." For n = 2 there are energy values Etil) in .tt"';b2
and E1il) in £~b2. They correspond to the eigenvalues
E7~ = -Ri'.i1 + i) = -68.0 eV
of Hoo and
E12 = - Ri'.e . 1 - R" . i = - 57.8 eV
of Ro. The experimentally measured values are
E I2 (0) - Etl(O) = 19.8 eV or E 1iO)= -59.2 eV,
Et2(0) - Et1(0) = 20.6 eV or Et2(0) = -58.4eV
for 1= 0, and
E I2 (1) - Et1(0)= 20.9 eV or E I2 (1) = -58.1 eV,
Eiil) - Eil(O) = 21.2 eV or Ei2(1) = -57.8 eV
for I = 1. Thus for n = 2 the energy values are already closer to the eigen-
values of H oo , as we expect from our general discussion above. For n =
3, 4, 5, ... the agreement improves further.
Figure 2.3 gives all known (1971) discrete energy levels of the helium atom.
This energy spectrum is in agreement with our discussion above. To summa-
rize our findings for the energy spectrum below E 100' there is one distorted,
shifted, and split hydrogen like spectrum of states in the space £~b2, and there
is one distorted, shifted, and split hydrogenlike spectrum-without the lowest
level-of states in the space £~bl. According to (1.23) the states in £~b2
are states with total spin zero, called singlet states; the states in £~b2 are
states with total spin one, called triplet states. As we shall explain below,
there are practically no transitions between the singlet and triplet states of
the helium atom. (The 2 3 S state is a quasistable state with a very long lifetime.)
Therefore the splitting of the energy diagram into two parts had in earlier
days lead to the hypothesis that helium was a mixture of two elements,
"parahelium," described by £~b2® PAs=o, and "orthohelium," described
by £~b2 ® PAs = 1.
XU Selection Rules and Singlet-Triplet Mixing for the Helium Atom 297
For parahelium the total angular momentum is equal to the total orbital
angular momentum L = I because the total spin S = O. Therefore the in-
teraction caused by the spin HI cannot split the energy level E7il), but can
only cause a small shift of the same order of magnitude as the fine structure
in the hydrogen atom, which has been calculated by approximation methods
(singlet).
For orthohelium the total spin S = 1. Therefore the direct-product
vectors
(2.45)
in A1- ® 9lS~ 1 are not eigenvectors of the total-angular-momentum op-
erators J2 and J 3' where
(2.46)
To obtain eigenstates of the total angular momentum we have to couple
the orbital angular momentum and the spin in A1 - ® 9ls~ 1 according to
the rules of Section V.2. Let vIt' denote the subspace of A1- c J't'~b2 with
orbital angular momentum I; it is the space spanned by Yn1l3 with a fixed value
of I. The eigenspaces vIt) of total angular momentum j are then given by
{ vltj~'+1 -:e vltj~' -:e vIt)~'-1 if 1"# 0

vIt' 'X' 9lS= 1 = . Q7 Q7 ' ( 2 . 4 7)
'CJ vIt J ~ 1 if I = O.
So we see that for orthohelium, except for the S states, all orbital-angular-
momentum spaces are actually a triplet of total-angular-momentum spaces.
Therefore the spin perturbation HI of the Hamiltonian H = H 0 + HI
splits the P (l = 1), D (l = 2), and F (l = 3) energy levels into triplets of
fine-structure levels, each belonging to an eigenspace of total angular
momentumj = 1+ 1,1,1- 1.
These results agree with the experimental energy spectrum: parahelium
consists only of singlet terms, whereas the orthohelium consists (except for
1 = 0) only of triplet terms (these are so close to each other that they are
shown by one line in Figures 2.2 and 2.3). To avoid labeling the states by their
symmetry properties under permutation ofthe orbital part (ortho- and para-),
one also calls the S term of the orthohelium a "triplet" term, Js 1, though it
consists in fact of only a singlet.
XI.3 Selection Rules and Singlet-Triplet Mixing for the

Helium Atom
The operator of the dipole moment for the helium atom is

(3.1)
where d is a vector operator with respect to the total angular momentum
operator
(3.2)
and also with respect to the total orbital angular momentum

Li = Lli ®I +I ® L 2i · (2.5)
Thus it obeys the commutation relations
[J i , dJ = iEijkd k (3.3)
[Li' d j] = iEijkd k (3.3')

As a consequence, one obtains by (V.3.7) for dipole radiation the selection
rules
J ~ J - 1, J, J + 1 (3.4)
L ~ L - 1, L, L+l (3.5')
The physical states are total angular momentum eigenstates, therefore the
selection rules (3.4) hold strictly for physical states. The selection rule (3.5')
holds to a very good approximation, precisely to the extent to which HI
commutes with L2.
The parity operator for the two-electron system is the direct product of the
parity operators in the one-particle subspaces:
Up = Up! <8l U p2 . (3.6)
Therefore, for the vectors (2.41) and (2.42), one obtains
Upl/ln1l3 = (-I)LI'/Ol/ln1l3' UpYn1l3 = (-I)LI'/IYn1l3 (3.7)
because L = I below the first ionization threshold. As (3.1) is a proper vector
UpdU;1 = -d,
it follows by the arguments of Section V.4 that the transition L ~ L is
forbidden, so that with (3.5') one has the selection rule
L ~ L + 1, L-l. (3.5)
A further selection rule follows from the fact that
Cd, IP 12] = o. (3.8)
For every vector
(3.9)
and
(3.10)
it follows as a consequence of (3.8) that
dy = y' E £~b2 ® {s: , (3.11)
dl/l = 1/1' E £~b2 <8l {s~ . (3.12)
Therefore,
(1/1, dy) = o. (3.13)
Consequently, there are no transitions between singlet and triplet states (the
same arguments as above hold for the quadrupole and higher multi pole
transitions, too). As already mentioned in Section XI.2, this had led to the
hypothesis that helium is a mixture of two elements-parahelium, possessing
only singlet states with an energy diagram given by the left part of Figure 2.2,
and orthohelium, possessing only triplet states with an energy diagram given
by the right part of Figure 2.2. However, one then observed transitions of very
weak intensity between energy levels of para- and orthohelium. These inter-
multiplet transitions are due to "singlet-triplet mixing"3 caused by HI' We
shall describe the eigenstates of total angular momentum in the spaces
Yf~b2 ® g?S= 0, Yf~b2 ® ~s= 1, which are obtained by coupling the spin and
orbital angular momentum, according to the rules of Chapter V. For
YfO:b 2 ® g?S=0, we obtain:
t/l~(~J = t/ln1l3=h ® L IS~>1Is~>2<tS~,ts~IOO> (3.14)

S3 5 3
where t/ln1l3 is given by (2.41) with (2.37) and the second factor on the right-
hand side of (3.14) is the same vector as the one given by (1.18).
For Yf~b2 Q9 ~s= 1 the total angular momentum eigenstates are given by
y~(3 = L Yn1l3
1)s3
Q9 (~, ItS~>1Its~>2<tS~, tS~llS3»
53S3
<113 Is 31J J 3>, (3.15)
where Yn 1l 3is gIven by (2.42) with (2.38). For a fixed n and I, y~(3 is a basis in the
spaces .4{J on the right-hand side of (2.47).
According to the considerations in Chapter VIII on perturbation theory,
the eigenstates of H = H °+ HI are in first order given by
.i,1h = ./,110
'I'nl 'I'nl
+ '\' (
L... E+ _ E+
1 ./,J1,<./,.I1)
'I'n'l' 'I'n'l'
1H 1 1./,11»
'I'nl
n'I' nl n'I'
n' =l=n
1'*1
(3.16)
and by
(3.17)
3 Mixing is the common term used for this kind of phenomenon, which is a somewhat
misleading nomenclature as compared with the meaning of the word mixture of states in
quantum mechanics. The singlet-triplet mixed states are not mixtures of states but rather
linear combinations of state vectors.
where E~ = E:1(I) and E;;j = E;;1 (I) are the eigenvalues (2.43), (2.44) of H 0
in the singlet and triplet states, respectively.
In (3.16) and (3.17) we have already taken into account that [H, JJ =
[H l' JJ = O. From parity conservation,
[H, Up] = 0 (3.18)
and, consequently, [HI' Up] = 0 then follows according to (3.7):
<t/J~~~IHllt/J~(3) =0 unless (-1)1' = (-lY,
(3.19)
<y~~~ IH lIY~(3) = 0 unless (-1)" = ( -1)',
and
(3.20)
In the perturbation series of (3.16), (3.17), only states of AI + or AI _ with

energy below the first ionization threshold have been included. This should
suffice for the qualitative considerations that we wish to present here. 4
The matrix elements <Y~~~ IHI IY~f 3) and <t/J~~~ IHI It/J~f 3) give the fine
structure contributions, i.e., the splitting of the terms 3(L)J into three energy
levels and a small shift of the energy levels of I(L)J. Calculations show (Bet he
and Salpeter, 1957, Section 40) that the J = 1 and J = 2 levels happen to
have almost the same energy.
The matrix elements <t/J~~~ IHI IY~f 3) give the singlet-triplet mixing, which
we shall discuss now.
H l' like every observable commutes with the permutation operator IP 12'
which represents the transposition ofthe two electrons. However, if we write
IP 12 as a direct product of an orbital part IPW and a spin part IP~ 2 according to
the representation (1.23)
(3.21)
then
(3.22)
i.e., HI contains contributions that are not symmetrical in the orbital operators
and spin operators separately. An example of such a contribution is the
interaction between the spin magnetic moment and orbital magnetic moment
of each electron which would, according to (IX.3.23), be given by
(3.23)
with
(3.24)
HI is only one term in HI : other terms like the relativistic mass correction term
4 For more details, see Bethe and Salpeter (1957), in particular Chapter lib.
and the interaction between the spin magnetic moment of one electron and the
orbital magnetic moment of the other are of the same order of magnitude. As
we want to discuss only the principle of singlet-triplet mixing, we will
restrict our discussions to (3.23).
In order to calculate the matrix element <Y~!L'IHllt/1~J3), we use (3.14)
and (3.15) and obtain
<Y~~~IHllt/1~J3) = L L C<Yn'I'llI C(1)L(1)I t/lnJh)<J J 311' 13 Is)(1s3I s3 s3)
X <s3IS(1)153)<s~ls~)<S3 s~IOO)
+ <Yn'I'1 31C(2)L(2) It/I nJJ)<J J 311' 13 IS3>0 s31 S3S~>

x <s~ IS(2)ls~> <s3Is3> <S3S~ 10 0» (3.25)
Changing in the second term the summation indices s~ ~ S3' s~ ~ 53 and
making use of the property of the Clebsch-Gordan coefficients <tS3 ts~ 11s 3>
= <ts~ t S31 1s 3); <tS3 ts~ 100> = ( -1) <ts~ tS31 00) one obtains
L L <J J 3 11' 13 1s 3>(1s3I s3 s~)<s3ISls3)<S3 s~IOO>
1353 5'3 5 3
53
x «Yn'I'llI C(l)L(l)I t/lnJJ) - <Yn'I'I,1C(2)L(2)I t/lnJJ3»' (3.26)
Since t/I nJJ3 is symmetric and YnTll is antisymmetric under 1P1r~ one has
<YnTI,1 C(2)L(2) It/lnJJ3> = - <YnTI l IC(l)L(l)lt/lnJJ)' (3.27)
To calculate these matrix elements, one uses (2.41) and (2.42) with (2.37) and
(2.38) and obtains four terms. Two of these terms are zero unless I' = 0, J = 0,
and one ofthem is zero unless l' = J. In order to reduce the number ofterms,
we shall restrict ourselves to the case l' ¥= J = 1 and obtain for this case
<Yn'l'l, IC(1) L(1) It/lnJJ) = - i<n' l' 131 u- t(n'l')C(i) L(i) U +(nJ) In J J 3>i
= -i<n'-l' 13 IC(i)L(i)In+ J J 3); (3.28)
where i can be either 1 or 2 and U-(n'/') and U+(nJ) are here the restrictions
of the unitary operators of (2.42) and (2.41) to ~rb. All we need to know of
them here is that they are scalar operators (because W is a scalar operator)
and do not change the value of l'13 and J J 3' Thus the new vectors
(3.29)
are again eigenvectors of L 2 and L3 with the same eigenvalue. Inserting (3.28)
and (3.27) into (3.26), one obtains
<Y~!fIHllt/1~J3>
= L L <JJ3ILI31s3)<ls3Its3,ts~><s3IS(i)ls3)<ts3,ts~100)
l353 5353
53
(3.30)
302 XI Two-Electron Systems- The Helium Atom
where C(i) is an L(i)-scalar operator. LU)K' K = 0 ± 1, are L(i) vector operators.

Thus
i<n'-LI3IC(i)L(i)Kln+J J 3); = <J J 3 1KILI 3><n' -LIIC u)L(i)lln+ J>. (3.31)
The reduced matrix element can also be further calculated using (Y.3.10):
(3.32)
>
Further evaluation of the quantity <n ' - I C(i) II n + J requires actual calculation
using perturbation theory with W as the interaction Hamiltonian or other
approximation methods for the determination of In± 113> and then cal-
culation of the matrix element <n'-II C(i) I n + >.1 using (3.24). We shall not do
this calculation here but just remark that these quantities are of the same
order of magnitude as the fine-structure splitting terms which are given by
<n'-IIC(i)lln- ).
Inserting (3.31) and
i<s~ISU)Kls3)i = <t s3 1 Klts~>Al (3.33)

into (3.30), one obtains
<Y~!lIHllljJ~(3) = L L L L <J J 3 1 KILI3><tS~,ts3100>
(3.34)
It is now a straightforward though slightly tedious calculation using the
symmetry properties (V.2.42)-(V.2.47) and orthogonality relations (V.2.14),
(V.2.15) of the Clebsch -Gordan coefficients to calculate (cf. Problem 3) that
<Y~!lIHllljJ~(3) = (-l)J-L+lj2b LJ JJ(J + l)<n ' -IIC(i)lln+)J' (3.35)

Therewith we have expressed the matrix element of the spin orbit interaction
between singlet and triplet states in terms of a quantity which (can be only
calculated numerically) is different from zero and of the same order as the fine
structure splitting term.
Inserting this quantity into (3.16) and (3.17) we see that the eigenvectors of
the operator H = H 0 + HI in first order, y~{3 and !jJ~{" are not pure triplet
and singlet states, but have a small component in the singlet and triplet space,
respectively. Thus, an orthohelium state y is mainly triplet y with a small
singlet component
y = }' + fljJ, If I ~ 1, (3.17')
and the parahelium state !jJ is mainly singlet IjJ with a small triplet component
!jJ' = 1jJ' + fyi, (3.16')
The dipole matrix element between para- and orthohelium is therefore
(3.36)
XI.4 Doubly Excited States of Helium 303
and even though the first term is zero according to (3.13), this matrix element
is different from zero due to the "singlet-triplet mixing."
We had mentioned above in Section VILI that it was impossible to verify in
the spectra of one-electron atoms whether the energy values were the ex-
pectation values of the energy operator in a state which is not an energy eigen-
state or the eigenvalues of the energy operator. From the helium spectrum
it follows that the latter is the case; the physical para- and orthohelium states
are eigenstates of H = H 0 + HI and not pure singlet and triplet states with
the energy values being the expectation values of H between these pure
singlet and triplet states. The existence of the transitions between para- and
orthohelium states is an experimental verification that the physical states are
eigenstates of the energy operator. In Chapter XII we will see that for theoreti-
cal reasons all stationary states have to be eigenstates of the energy operator.
XI.4 Doubly Excited States of Helium
The energy levels E In,(I) of the helium atom are experimentally very well
known. They have been measured by light spectroscopy, i.e., by measuring
the wave number (or frequency) of the emitted or absorbed light in a transi-
tion between two energy levels according to
The energy levels have also been measured in energy-loss experiments as

described in Section 11.4, and as depicted schematically in Figure 11.4.1.
The collision chamber was filled with helium gas, and the intensity of the
electron current was measured as a function of the energy lost in collisions
with the helium atoms. The results of this experiment are shown in Figure
4.1, where the intensity I has been plotted as a function of the energy loss E.
According to the interpretation of this experiment in Section 11.4, the energy
lost by the electrons is used to excite the atom and cause the transition from
the ground state 11 S to an excited state; therefore the intensity shows maxima
at values of energy loss that agree with the energy difference between the
ground state and one of the states excited by this experiment. The excitation
energies that are observed in this experiment are tabulated in Table 4.1,
column 3, and are compared with the energy of the n lp levels measured by
light spectroscopy (column 2); the first column gives the energy levels that
participate in the transition. Not given in the table are the bumps between
42-e V and 46-e V energy loss. These bumps are the result of two successive
excitations (double scattering); the electrons in the collision chamber
scatter inelastically twice, losing energy to the transition 1 1S ~ 2 1P and
to the transition 1 1S ~ nip, or two times to the same transition 1 1S ~ 3 1P.
From Figure 4.1 we see that there is a very weak transition 1 1S ~ 2 1S,
and that the transitions of high intensity are the 1S ~ 1P transitions (dipole
Table 4.1 Excitation energies of helium.
Excitation energy (eV) measured in
light spectros- energy-loss

Transition copy experiment experiment
IIS-2Ip 21.212 21.213

31 P 23.081 23.084
4 1p 23.736 23.739
51 P 24.039 24.034
61P 24.205 24.209
71 p 24.304 24.308
81 P 24.369 24.374
9 1P 24.413 24.419
IO l p 24.445 24.451
II l p 24.469 24.474
121 P 24.487 (24.493)
2 1S 20.610 20.612
2s2pl p 60.123 60.120
2s3pl p 63.651 63.651
2s4plp 64.462 (64.450)
transitions). No transition between the liS ground state and the ID, IF, ...
states or between the 1 I S state and the triplet states are observed, which
agrees with what one would expect for dipole excitations.
Figure 4.1 shows that there are further bumps above 60 eV; these bumps
look different from the lower-lying bumps and have a typical asymmetric
profile. They correspond to energy levels that are well above the ionization
energy for one electron. Energy levels with such a high energy above the
ground state level have also been observed as spectral lines in the optical
(ultraviolet) absorption spectra. Figure 4.2 shows a photograph of the
absorption spectrum obtained for helium in the region 160-215 A. The
observed discrete spectral lines are superimposed upon a continuous ab-
sorption background. The lowest-lying discrete spectral line in this region
is at 206.21 A, corresponding to an energy of 60.1 e V; this agrees well with
the position of the first bump of this kind in the energy-loss spectrum of
Figure 4.1 (cf. also Table 4.1). It is the lowest-lying line of a single long series
of discrete spectral lines of outstanding intensity that converge to the energy
value 65.4 eV. (Some of these lines have corresponding bumps in the energy-
loss spectrum of Figure 4.1.) Three fainter lines indicate the existence of a
second series that also converges towards this energy value. A number of
fainter lines can be seen in Figure 4.2 lying to shorter wavelengths from the
most prominent series. These lines are grouped into series having as limit
points the energy values 72.9 eV and 75.6 eV. Thus we have a series of series of
spectral lines. As these lines are observed in the absorption of electro-
magnetic radiation by helium atoms in the ground state, we can safely
assume that (at least for the high-intensity lines and probably for the fainter
/0'
l'S - n 'P
He'
10)
1101 (1 '5 - 2'P!

' /1 'S - n 'Pi x
:"..
lslp'P lsJp'P 2sl,p'p
I I
-
o
o
s::
10
r::r
~
tTl
><
g.
("0
doub le-sca ll.:ring 0-
double-excilalion [FJ
§:
("0
20 22 2' '~?----- " '6 60 62 6' eV V>
E- o
..,
Figure 4.1 Energy-loss spectrum of helium. [From H, Boersch, J. Geiger, B. Schroder, Abhandlungen der Deutschen l:
g.
Akademie der Wissenschaften, Berlin (1967), No. 1, p, 15, with permission.] ;::.
3
Vol
o
Vl
79.0 75.6 72.9 65.4 E - Etl (Q = 0) (eV) w
o
~
..... 1 1 1 I
"'-1- 1- T
-><
--l
170 E 210~ ~
o
+ m
r>
...,~
o
::s
en
'-<
V>
rt
3
V>
!
--l
;:r
(l)
::r:
~
c
3
;l>
/353 P Ip~ o
, !il I 3
l··l/ 1 5P,23-
j II 1454P 'P,O
t
++
t t t
He He+(n=4) H:(n=3) H/(n=2)
Figure 4.2 The absorption spectrum of helium between 160 and 215 A showing the many resonances due to the existence of
two-electron excitation states in neutral helium. The photograph is a positive print (black is absorption). The white portion
of each resonance indicates a reduced-adsorption zone or "window" in the continuous photoionization background
absorption. [From R. P. Madden, K. Codling: Astrophysical Journal 141, 364 (1965), with permission.]
XI.4 Doubly Excited States of Helium 307
lines too) they correspond to dipole transitions and that (according to the
selection rules for dipole transitions) the corresponding excited energy
levels belong to singlet states with orbital angular momentum 1, i.e., are 1 P
states.
The threshold for double ionization of the helium-i.e., the energy
value Eoooo at which both electrons have just dissociated from the helium
nucleus and the physical system is in a state consisting of the nucleus and
two electrons, both with relative energy O-lies 79.0 eV above the ground
state. We stated this experimental fact before by saying that the ground-
state energy Eil = -79.0 eV. [The energy scale in Figure 4.2 gives E(levei)
- Eil']
The observations in Figure 4.2 indicate that one has a series of series of
1 P energy levels, and it appears that the limit point of this series of series lies
79.0 eV above the ground-state energy, i.e., at the threshold for double
ionization. The limit points for each single series are then seen to agree with
the energy values Enoo as calculated from (2.33):
Enoo - Etl = -RHe~

n
+ 79.0eV = (-54.4eV)~ + 79.0eV;
n
(4.1)
hence
E 100 - Eil = 24.6 eV,
E 200 - Etl = 65.4 eV,
(4.2)
E300 - Etl = 72.9 eV,
E4cr - Eil = 75.6 eV.
Thus the observations in Figure 4.2 show that the discrete 1 P energy levels
converge to the nth ionization threshold of the helium atom, i.e., to the
energy value at which one electron has the principal quantum number nand
the other has just dissociated itself from the helium atom. These experi-
mentally measured 1 P levels above the first ionization level are depicted in
Figure 2.3. Also depicted there are all the other known (1971) discrete
energy levels of the helium atom. We see that the energy-level diagram
below the second, third, etc. thresholds is qualitatively very similar to the
energy diagram below the first threshold.
A rough explanation of the discrete energy levels above the first ioniza-
tion threshold is now very simple: 5 Whereas the energy levels below E 100
correspond to states in which one of the electrons is always in the ground
state (n = 1), the energy levels below E noo ' n > 1, correspond to states in
which one of the electrons is in the state with principal quantum number nand
the other is in a state with n' ;:::: n. Both electrons are excited and the atom is
5 Extensive theoretical calculations of these energy levels by various approximation methods

have been performed and agree with the observed values; see J. Macek, J. Phys. B 1, 831 (1968)
and references therein. See also U. Fano: Doubly Excited States oIAtoms in Atomic Physics,
edited by V. W. Hughes et al., Plenum. New York (1969).
said to be doubly excited. In the approximation where the interaction W

between the two electrons is neglected, these are the states that correspond
to the vectors (2.15 ±) with n' ~ n > 1. For the case n = 2, (2.15 ±) reads
t/lg1l3.n'1'I3 = fi (12[[3)1 @ In' I' 13)z + In' I' 13)1 @ 12 [[3)Z)' (4.3+)
ygll n'l'I'3=
3, y2
~(12/13)1@ln'1'13)z-ln'1'13)1@12113)z), (4.3-)
where 1 can take the two values I = 0 and 1 = 1; i.e" we have the singlet
states t/lg lb. n'I'13 and t/lgoo, n'1'I3 and the triplet states yg 1l3. n'l'l, and ygoo, n'l'13'
If we want to explain the I P levels, we have to couple the angular momenta
1 and I' to obtain eigenvectors of total orbital angular momentum L Z =
(LI + Lz)Z with eigenvalue L(L + 1) = 1(1 + 1). As L can be any of the
values I + 1', 1 + I' - 1, ... , II - /' I, we can obtain L = 1 in the following
cases:
2snp 1= 0 I' = 1 L=I+1' n= -1
2pns 1 = 1 1'=0 L=I+1' n= -1
(4.4)
2pnp 1 = 1 I' = 1 L=I+I'-1 n= +1
2pnd 1 = 1 I' = 2 L=I+/'-2 n= -1
where n is the parity, n = (-IY( -IY'. Therefore one has three possibilities
to obtain I P states with parity n = - 1 as linear combinations of the states
(4.3 ±):
(n = 2, I = 0; n', l' = 1), (4.5p)
(n = 2, I = 1; n', /' = 0), (4.5s)
(n = 2, I = 1; n', I' = 2). (4.5d)
All of these linear combinations are eigenvectors of Hoo with the same
eigenvalue £~n~' The energy shift due to the interaction of the two electrons
depends not only on n' and L but also on I and l'; one would therefore expect
three different energy values for every value of n' (except for n' = 2). Thus
one would expect there to be three series of energy levels converging to the
£zoo limit. Figure 4.2 shows only two series, one with a very high transition
probability from the 1 IS ground state and the other with a very low one.
This indicates the existence of an approximate selection rule that cannot
be explained on the basis of the assumption that the spaces spanned by (4.5p),
(4.5s), and by (4.5d) are good approximations to spaces of physical states.
It is believed that linear combinations of (4.5p) and (4.5s) span, to a good
approximation, the spaces of energy eigenstates that belong to the two ob-
served series,6 and that (4.5d) corresponds to a still fainter and therefore yet
unobserved series.
6 J, W. Cooper, U. Fano, and F. Prats, Phys. Rev. Left. 10,518 (1965).
Problems 309
The situation becomes even more complicated for n = 3, 4, .... With

similar arguments the 3 P series can be discussed.
We conclude the discussion of the doubly excited states with a remark
concerning the broad profile of the spectral lines corresponding to these
states, which is shown in Figure 4.1 and is also visible in the optical spectra.
These energy levels lie above the first ionization threshold. Therefore, in
addition to the doubly excited states with n = 2, n' ;;?: 2 there are states with
the same energy as the energy of the doubly excited states which belong to
the system that consist of the He + ion in the ground state n = 1 and one
electron (He+ + e). The system He+ + e can take any possible energy
value E ;;?: E 1 <Xl' because the relative energy between He + and e can take
any value, i.e., the energy spectrum of He+ (n = 1) + e is continuous above
E 1<Xl' Therefore each energy value of the doubly excited states coincides
with one of the energy values in the continuous spectrum of the states
of the system He+ (n = 1) + e. Thus there are two different states with the
same energy: the discrete state n = 2, n' ;;?: 2, and the state of the continuum
with n = 1 and a relative energy between He + (n = 1) and e. These two states
mix and interact with each other, as a result of which the doubly excited
state cannot persist and easily decays into the state of a He + (n = 1) and an
electron. The lifetime of such doubly excited states is therefore much smaller,
and the spectral line has the typical broad profile. Such states are often called
resonances.
For the doubly excited states with n = 3, n' ;;?: 3, there exist two continuum
states with the same energy, namely He+ (n = 1) + e and He+ (n = 2) + e.
For n = 4, n' ;;?: 4 there are three, and so forth.
Problems
1. Show that the symmetric and antisymmetric parts of the total space i) are given by
.;t'~ = (.;t'~b2 ® ~~) EEl (.;t'~b2 ® ~~),
.;t'~ = (.;t'~b2 ® i~) EEl (J'f"!b 2 ® i~),
respectively.
2. Check whether HI given by (3.23) commutes with the operator of total orbital
angular momentum squared, L2.
3. Using the properties of the Clebsch-Gordan coefficients given in Section V.2,
perform the calculations that lead from (3.34) to (3.35).
CHAPTER XII
Time Evolution
In this chapter the basic assumption on time evolution is introduced. The

Schrodinger picture, the Heisenberg picture, the Dirac picture, and their
relations are discussed. Time-dependent external forces are only briefly
mentioned at the end of the chapter.
XII.1 Time Evolution
Until now we have not mentioned time and have completely ignored the
time evolution of the state of a system and the time development of an
observable. We have restricted ourselves to properties of systems measured
at one instant. In general, one should specify not only the physical values
obtained in a measurement but also the time at which the observation is
made. In this chapter we consider the time evolution-the dynamics-of
physical systems. Although the notion of one instant of time makes as little
sense physically as the notion of a definite position (measurements always
require a finite amount of time in the same way that they always require
a finite amount of space), we make the idealizing assumption that time is a
continuous real parameter t that labels a sequence of states. (Note that we
have called time a "parameter" as opposed to an "observable.") That is, we
make the idealizing assumption that measurements can be made at a succes-
sion of arbitrarily small time intervals, and that at any instant of time the
specification of a complete set of observables is possible.
Let us denote by W(to) the state of a system that is prepared at the time
t = to' We then expect that W(to) will uniquely determine another state
310
XII. 1 Time Evolution 311
W(t l ) at any later time t = tl > to' During the evolution the system may be
subject to external influences, so that the manner of evolution will depend
explicitly upon the interval from to to t I -its starting time to as well as its
length t I - to. We first consider systems that evolve undisturbed by external
influences (isolated systems). For such systems the evolution
(1.1)
does not depend upon to but only upon the system itself, the initial state
W(to), and the length, I = t I - to of the time interval from to to t I' Our
intuitive understanding of the time evolution suggests that we impose the
following requirements on the map (Ll):
1. Expression (Ll) is a linear transformation of the set of states, con-
tinuous in a certain sense: If WI(t O) ....... W I(t 1 ) and if W 2(t O) ....... W 2(t l ),
then for a, bEe,
a WI (to) + b W 2(t O) ....... a WI (t d + bWit d.
2. Time evolution forms an additive semigroup; i.e., if W(t o) ....... W(t o + 'I)
for 'I = tl - to and if W(t l ) ....... W(tl + '2) for '2 = t2 - t l , then
W(t o) ....... W(t o + '1 + '2)'
Requirement 1 expresses the physical statement that first mixing the states
at time to and then letting the system evolve in time leads to the same results
as first letting the states evolve separately in time and mixing them afterwards
at time t I; the continuity of the transformation expresses the intuitive
statement that time changes continuously. Requirement 2 is intuitively
obvious; it is a statement of the fact that letting a physical state evolve first
for a period of length 'I
and then letting the resultant state evolve for a
period of length '2 leads to the same result as letting the original state
evolve for a period of length, 1 + '2' Requirements 1 and 2 are very general
and must be satisfied by irreversible processes also.
We are here concerned with the quantum mechanics of reversible processes,
and for such processes we make the additional requirement
3. The time evolution (1.1) is given by a unitary operator U( ,) satisfying
W(to) ....... W(t l ) = Ut('I)W(tO)U('I)' (1.2)
ute,) = U- I (,), (1.3)
U(O) = I, (1.4)
U- I (,) = U( -c), (1.5)
U('I + '2) = U('I)U('2) (- 00 < 'I' '2 < + ex)), (1.6)

U(,) is a continuous operator function I of the param-
eter " (1.7)
1 See appendix to this chapter.

312 XII Time Evolution
Equation (1.5) follows from 0.2) and (1.3)-specifically from

U(r 1)W(t 1)U-1(r 1) = W(to) (1.2')
-and from the assumption that (1.2) is also valid for evolution backwards
in time, i.e.,
(1.8)
Equation (1.6) follows from requirement 2, which, when the time evolution
is described by a unitary operator as in (1.2), is expressed as
(1.6')
Equation (1.6) follows for r 1 = t1 - to' r 2 = t2 - t 1. The linearity of the
transformation (1.2) is obvious. The continuity of the transformation (1.1) is
expressed by (1.7). We shall not give here the mathematical definition of a
continuous operator function, but mention only that the formal operations,
employed below, of differentiation, Taylor series expansion, etc. 2 can be
rigorously defined. A continuous operator function satisfying (1.6) is called a
one-parameter group of operators.
We differentiate U(r) with respect to r and define
A == dU(r) I . (1.9)
dr ,=0
A is called the infinitesimal operator or generator of the operator U(r) of

time evolution. U(r) is not only "differentiable," but is differentiable an
arbitrary number of times. We can calculate all "derivatives" using (1.6).
Differentiating (1.6) with respect to r 1 ,
dU(r + r 1 ) d(r + r 1 ) _ dU(r 1) U(r)
d(r + r 1 ) dr1 - dr1 '
and evaluating the resultant expression at r 1 = 0, we obtain
dU(r) = A ()
dr U r.
We differentiate this again and obtain
d 2 U;r) = A dU(r) = A. AU(r) = A 2 U(r).

dr dr
We may continue to differentiate, obtaining for the pth derivative
dPdU(r) = APU(r);
rP
use of (1.4) then gives
dPU(r) I _ AP
dr P ,=0 - .
2 See the appendix.

XII. I Time Evolution 313
The "Taylor series expansion" of U(r) is therefore given by

r2 rn
U(r) =I+ rA +- A2 + '" + - An + .... (1.10)
2 '. n., '
or, if we compare (1.10) with the series for eTA, we have
U(r) = eTA. (1.10')
In particular this means [according to our rule that f(A)la) = f(a)la)

for functions f(A) of A] that
for (generalized) eigenvectors Ia) of A corresponding to the eigenvalue a.

Taking the adjoint of (1.10) gives
r2
Ut(r) =I + rAt + 2! At2 + ... = e TAt • (1.11)
If we use (1.3) and differentiate Ut(r)U(r) = I, we obtain
d~:(r) U(r) + Ut(r) d~;r) = 0;

setting r = 0, we obtain
At = -A. (1.12)
An operator that fulfills (1.12) is called skew-Hermitian. As r is the time, A
has the dimensions of frequency, i.e., of energyIii. The operator
H = -iliA = -iii dU(r) I ' (1.13)

dr T=O
often also called the generator of time evolution, is therefore a Hermitian

operator with the dimension of energy. H is the Hamiltonian operator or
energy operator of the system, and, from (1.10'),
(1.14)
The precise form of H depends, of course, upon the particular physical
system under consideration, and how it is found in particular cases has been
discussed previously. H is an element of the algebra of observables of the
particular system, and its properties are determined by the defining algebraic
relations of the particular physical system. We remark that there is a one-to-
one correspondence between U(r) and H; one can therefore consider H as a
quantity derived [using (1.13)] from the more fundamental time-evolution
operator U(r), or one can consider H as the more fundamental quantity
and the time-evolution operator U(r) as a quantity derived [using (1.14)]
from H. It is in the spirit of our preceding formulation to consider H, an
element of the algebra of observables, as the fundamental quantity, and to
consider U(r) as the derived quantity.
We formulate the result of our preceding plausibility argument as the

basic assumption of quantum mechanics for time evolution-the basic
dynamical law for quantum-mechanical systems:
Va (The Schrodinger Picture). A (conservative) physical system has a

generator H of time translation which is a Hermitian element of the algebra
of observables and which is characteristic of the physical system. The time
evolution of the state of the physical system is given by
WU) = ut(t)Wo U(t), (1.15)
where Wo is the state of the system at some initial time to = 0 and where
U(t) is given by
(1.16)
According to the axiom II, the expectation value of an observable A is
given by
<A) = Tr(AW). (1.17)
The observable that is represented by the operator A is defined by the pre-
scription for its measurement, i.e., how to set up a measurement apparatus
and how to take the measurement. This prescription is given for all time and
does not change with time. The experimentally measured value changes in
time; if we make a measurement at a time to, the expectation value <A >r = to
will, in general, be different from the expectation value <A)t=tl at a different
time t 1. This experimental fact is represented by a description of the change
of the state Wet) according to (1.15). Inserting (1.15) into (1.17), we obtain
<A>r = Tr(A Wet»~ = Tr(AUt(t)Wo U(t». (1.18)
This can be written
(1.19)
since the trace of a product of operators is invariant under a cyclic permuta-
tion of the order of the factor operators. We can now define a new operator
A(t) by
AU) = U(t)AUt(t) (1.20)
and write (1.19) as
<A>r = <A(t» = Tr(A(t)Wo)· (1.21)
In (1.21) the change of the experimentally measured value <A>r with time is
described in a form that can be interpreted in the following way: The state
of the system is described by an operator Wo; this statistical operator Wo
does not change with time, i.e., the whole history of the state of the physical
system is described by the one operator Wo. The observable develops with
time according to (1.20), i.e., the prescription for the measurement must
include instructions as to when to take the measurement, because the ob-
servable-the experimental setup-is different for different times. Thus
instead of the basic assumption Va we can formulate the time evolution by

the following basic assumption:
Vb (The Heisenberg Picture). A (conservative) physical system has a

generator H of time translation which is a Hermitian element of the algebra
of observables and which is characteristic of the physical system. The time
development of every observable A(t) of the physical system is given by
A(t) = U(t)AUt(t), (1.22)
where A describes the observable at some initial time to = 0 and where U(t)
is given by
(1.23)
These two descriptions of the change in time of the physically measured
value (A)t are, of course, completely equivalent. In the form Va the change
in time is described as a change of the state of the system. The description of
the change in time in this form is called the" Schrodinger picture." In the
form Vb the state is held fixed and the change in time is described by the
change of the observable. The description in this form is called the" Heisen-
berg picture."
The time development of an observable as given by (1.22) is often written
in a different form, which is obtained from (1.22) by differentiating:
dA(t) = ~ HeitH/hAe-itH/h + eitH/hA(_ ~ H)e- itH /h

dt h h
i
= h (HA(t) - A(t)H),
or
. dA(t)
lh at = [A(t), H]. (1.24)
This is the" Heisenberg equation of motion." An observable is a constant

of the motion if it does not depend upon t. From (1.24) it follows that all
constants of the motion commute with H. In particular, H itself is a constant
of the motion. Furthermore, every member of any c.s.c.o. that contains H is
a constant of the motion.
Consider the particular observables momentum Pi and position Qi for a
physical system in which neither of them is a constant of the motion. It is
easily seen that [Pi(t), Q/t)] has its usual form for all times t > 0 given
that [Pi' Qj] = (hji)J i) at t = 0:
Pi(t)QP) - Q/t)Pi(t) = U(t)P i ut(t)U(t)Qj U\t) - U(t)Qj ut(t)U(t)P i ut(t)
= U(t)(PiQj - QjP;)ut(t)
_ t h _ h
- U(t)U (t) -:- Jul - -:- Jijl,
I I
I.e.,
(1.25)
We are now in a position to calculate the equations of motion for the

momentum and for the position and to show that these equations have the
same form as those for the classical quantities. For a system whose
Hamiltonian is given by
1
H = - PiP i + V(Q),
2m
where
we calculate that
1 1
[Qk' H] = 2m [Qk' PiPJ = 2m {[Qk' PJ, PJ
Inserting this into (1.24), we obtain

dQk(t) Pk(t)
(1.26)
dt m
This is the quantum-mechanical analogue of the classical relation between
momentum and the time derivative of position.
In classical physics the force is obtained from the potential function
Vex) by
fk = _ 8V(x).
8x k
In analogy to this we define the "force operator" by
i "3V(Q)"
Fk == - h [P k, V(Q)] = - - 8Qk (1.27)
where the last symbol is defined below. Inserting this into (1.24) for A(t) =
Pk(t), we obtain
dPk(t) 1 i
----;{t = ih [P k, H] = - h [P k, V(Q)] = Fk(Q(t)). (1.28)
This is the quantum-mechanical analogue of the classical relation between

the time derivative of the momentum and the force. Combining (1.26)
and (1.28) gives the quantum-mechanical Newton's equation,
m d2ir~(t) = Fk(Q(t)). (1.29)
[" 3V(Q)/3Qk" is a symbolic way of indicating the following pro-

cedure: Replace Qi by the numbers Xi in V(Q) [recall that this is just
the inverse to the way V(Q) was obtained from the corresponding
classical expression], differentiate Vex) with respect to Xk> and

replace Xi by Qi in oV(x)/ox k • That "oV(Q)/OQk" is in fact (i/h)
[Pk> V(Q)] can be seen in the following ways:
1. Write
V(Q) = C + CiQi + CijQiQj + ...
+ Cil'''inQi l ... Qin +"', (1.30)
where the Cil"'i n are symmetric, i.e.,
(1.31)
because the Qi commute with each other. Using the canonical
commutation relation (1.25), calculate that
because
= Cil"'i n L [P k , Qi)Qil ... Qip-lQip+l ... Qin

p=l
/I h
= C it .. ·i" L1 -;-bkiI'Qit'"
p= I
Qip-tQip+t '" Qin
where (1.31) has been used. The right-hand side of (1.32) is

identical to "aV(Q)/OQk'"
2. Calculate [P k , V(Q)] in the position representation:
(xl[P k, V(Q)]ll/I) = (x IPk(V(Q) I l/I» - (xlV(Q)P kI l/I)
h 0 h 0
= -;--(xlV(Q)ll/I) - V(x)-;--(xll/l)
I oX k I oX k
h 0 h 0
= -;--0 (V(x)(xll/l» - V(x)-;-;-(xll/l)
I Xk I uX k
= ~ oV(x) (xll/l). ]
I oX k
As the Heisenberg commutation relations do not change with time, all

algebraic relations that are derived from them are also valid at any time,
even if the operators in these algebraic relations do not commute with H.
That is, any algebra of observables whose elements are functions of Pi' Qj
does not change in time, even though some of the elements may not commute
with H.
A symmetry transformation of a set of elements is a transformation of the

elements that leaves the relations among them unchanged. Thus, the time
evolution of a physical system whose algebra of observables is derived from
the Heisenberg commutation relations is a symmetry transformation of the
algebra of observables. If time evolution is a symmetry transformation, then
the mathematical structure (in particular the algebraic relations) of the
algebra of observables does not change in time; this means that the physical
structure is indistinguishable at two different points in time. Our experience
shows that there are physical systems that have this property, and in fact it is
this property that defines the isolated physical systems. Thus isolated physical
systems do not age, an absolute value of time has no meaning for these
systems, and only time differences are accessible to measurement. Irreversible
processes do not take place in isolated physical systems defined as above.
These considerations lead us to another formulation of the basic assump-
tion V:
Ve. The time evolution of a conservative physical system is given by a

continuous symmetry transformation of its algebra of observables.
Let us return to the Schrodinger picture. The time evolution (1.15) can
also be expressed in differential form. Differentiating (1.15), one obtains
dW(t) i i
----;It = h (W(t)H - HW(t» = h [Wet), H]. (1.33)
We consider now the special case where the state Wo is a pure state, i.e.,
(1.34)
where A,I/Io) is the projection operator onto the one-dimensional space

spanned by the vector I!/J 0)' For this case (1.15) reads
Wet) = Ut(t)A'I/Io) U(t) = Ut(t) 1 !/Jo)<!/J o 1 U(t).
We define the vector !/J(t» by 1
I!/J(t» == Ut(t)l!/J o) (1.35)

and have
Wet) = Ut(t)A,I/Io) U(t) = 1 !/J(t» <!/J(t) 1 = A'I/I(t»· (1.36)
We obtain the differential form of (1.36) by differentiating (1.35):
:t II/I(t) = d~tt(t) 11/10) = - ~HUt(t)ll/Io) = - iH11/I(t).
This is the Schrodinger equation for the state vector II/I(t):

d
in dt II/I(t) = HII/I(t». ( 1.37)
For the Hamiltonian given by

1
H = 2m PiPi + V(Q),
the SchrOdinger equation in the position representation is obtained from this
by taking the scalar product with the position eigenvector Ix):
d 1
ih dt (xll/I(t» = 2m (XIPiPill/l(t» + V(x) (x II/I(t».
Using
(see Equation (11.7.58», the Schrodinger equation may be written in the

conventional form
d h2
ih dt I/I(x, t) = - 2m V 2 1/1(x, t) + V(x)l/I(x, t), (1.38)
where we have defined the time-dependent wave function

I/I(x, t) == (x II/I(t». (1.39)
Summarizing, the distinction between the Heisenberg picture and the

Schrodinger picture is the following: In the Heisenberg picture the observa-
bles are transformed by Vet) and the states are kept fixed, while in the
Schrodinger picture the observables are kept fixed and the states are trans-
formed in the "opposite direction," i.e., by vt(t) = V-let). The time depen-
dence of Tr(A W) is the same in both cases. Expressed in terms of vectors: In
the Heisenberg picture the eigenvectors la t) of the observables A(t),
A(t)la t) = ala t), (1.40)
are "rotated" by Vet):
V(t)laO) = lat),
whereas the state vectors 11/10) are kept constant. [Note that in (1.40) no
time dependence occurs on the part of the eigenvalue a. This is-easily shown:
A(t)lat) = A(t)V(t)laO) = V(t)A(O)laO) = V(t)a(O)laO) = a(O)lat).] In
the Schrodinger picture the eigenvectors Ia 0) are kept constant and the
state vectors are" rotated" in the opposite direction:
The time dependence of the physically measurable quantities representing

the probabilities is in both cases the same:
l(atll/l o)12 = I(V(t)laO), Il/Io)W
= l(aOIVt(t)ll/Io)1 2 = l(aOII/l(t»1 2.
A state is called a'stationary state if it does not change with time:

Wet) = vt(t)Wo Vet) = Wo = W. (1.41)
This terminology is justified by the fact that every observable measured in
this state gives a value independent of the time of the measurement; for from
(1.41) it follows that the expectation value of any observable A is
<A)t = Tr(AW(t» = Tr(AVt(t)WV(t» = Tr(AW) = <A\=o'
Stationary states are always mixtures of eigenstates of the energy operator H.
This follows from (1.41) in the form
or equivalently,
WH-HW=O. (1.42)
In particular, for a pure stationary state,
(1.42) reads
or
HI,I, ) = 1,1, ) <l/IoIHll/lo) (1.43)

'1'0 '1'0 <1/1011/10)'
Thus pure stationary states are represented by eigenvectors of the energy
operator. The time dependence of a stationary state vector is given by
(1.44)
Stationary systems are the systems that have only stationary physical states.
Stationary systems can therefore only be prepared in states that are mixtures
of eigenstates of the energy operator, i.e., in states that fulfill (1.42). And
pure physical states of stationary systems must always be represented by
energy eigenvectors. In the preceding chapters all physical systems con-
sidered were stationary physical systems, and we have now found an" explana-
tion" for the fact that these systems always appeared in eigenstates of the
energy operator.
To what extent a given physical system may be considered stationary
depends upon the desired accuracy of the description. If one ignores details,
atoms and molecules are stationary physical systems. If a measurement
process has prepared them to be in a particular state, which has to be an
energy eigenstate, then they will for "all time" remain in that state. "All
time" means fairly long on atomic scales: 10 - 2 sec (for infrared transitions)
to 10- 9 sec (for ultraviolet transitions) for the excited states of atoms and
molecules. If one wants to 'describe not only energy levels but also details of
the decay process of excited states, then one cannot consider the atom as a
stationary system. This will be discussed in later chapters.
The Heisenberg picture and the Schrodinger picture are the limiting cases
of a more general picture, which is called the" Dirac picture" or the" inter-
action picture." One writes the Hamiltonian operator H as the sum of two
parts,
(l.45)
How this splitting is performed, i.e., how one makes the choice as to which
part of H is to be H 0 and which part is to be HI' depends upon the particular
situation. In principle any part of H can be called H 0 and the rest then called
H l' In practice one chooses for H 0 an operator whose eigenvalues are known
or can easily be obtained. Often H 0 will represent the energy of two or more
combined systems when the interaction between the systems is ignored;
HI then represents the interaction energy.
Assuming that H is not explicitly time-dependent (a conservative system),
we define, in accordance with (1.16),
Vet) = eitHlh. (1.46)
The expectation value of an observable A at time t is then given by
(A)t = Tr(V(t)AVt(t)Wo), (1.47)
where A represents the observable at some initial time to = 0, and where Wo
is the state of the system at that initial time. If we define operators
(1.48)
and
VI (t) == V(t)vb(t) = V(t)e - itHoh, (1.49)
then (1.47) may be rewritten as
(A\ = Tr(V 1 (t)V o(t)AV6(t)V1{t)Wo)
= Tr(V o(t)AVb(t)VI(t)Wo V l(t»
= Tr(AD(t)WD(t», (1.50)
where
(1.51)
and where
(1.52)
Note the similarity between the time development (1.22) of A in the
Heisenberg picture and the time development (1.51) in the Dirac picture.
In the former the evolution operator Vet) = eitHlh is generated by the full
Hamiltonian H, while in the latter V oCt) = eitHolh is generated by only part
of the Hamiltonian, namely H o. In the extreme case that H 0 = Hand HI = 0
we have V oCt) = Vet) and VI (t) = I, so that (1.50), (1.51), and (1.52) re-
duce to the Heisenberg picture. Alternatively, there is a similarity between
the time development (1.15) of the statistical operator W in the Schrodinger
picture and the time development (1.52) in the Dirac picture. This similarity
is even more pronounced if
(1.53)
for then (1.49) becomes
(1.54)
In the case (1.15), V(t) = e itH / h is then generated by the full Hamiltonian
H, while VI (t) = eitHl/h is generated by the interaction part of the Hamilton-
ian, namely HI' In this extreme case H 0 = 0 and HI = H, we have V oCt) = I
and VI (t) = Vet), and we thereby regain the Schrodinger picture.
Suppose Wo describes a pure state,
( 1.55)
then we can define a vector
(1.56)
such that, according to (1.52), WD(t) is expressed as
(1.57)
Differentiating (1.56), one obtains for the most common case that (1.53)
holds:
d
ih dt 1!/I(t»D = H 1I!/1(t»D' (1.58)
Thus the vectors representing the state in the Dirac picture obey the
"Schrodinger" equation with the interaction Hamiltonian HI'
In the interaction picture the time dependence of the physically measurable
quantity <A\ is described by letting both observable and state vary with
time. The observables are rotated in one direction with a transformation
generated by part of the Hamiltonian, and the states are rotated in the op-
posite direction by a transformation generated by the other part (the inter-
action part) of the Hamiltonian. Expressed in terms of vectors: The eigen-
vectors at) D of the observable A,
1
are given by a rotation of 1 a 0) D by eitHo/h,
one obtains the state vectors I!/I(t» D by rotation of I!/Io) in the opposite
direction by VIet) (= e-itHt/h if [Ho, HI] = 0):
The time dependence of the probabilities is the same as in the case of the
Heisenberg and Schrodinger pictures:
l<atlt/lo)12 = l<aOIU t (t)lt/lo)1 2

= I<a 01 U6(t)U1(t) 1t/lo) 12
= ID<a tl t/I(t»D 12.
Isolated systems are an idealization, because a physical system cannot be
completely isolated from the rest of the world. However, in many cases these
idealizations give a sufficiently good approximation to the physical situation.
If the isolation between a physical system and some part of the rest of the
world is not negligible, one can enlarge the physical system by this part of
the rest of the world and then consider the combined system. If that is not
sufficiently isolated, one can continue to enlarge the physical system and will
ultimately end up with an isolated system to which the previous considera-
tions about time development apply. In many situations, however, this
procedure is not practical. The enlarged system might be too complicated
to be described by a sufficiently simple mathematical structure (if such a
structure exists at all). In such a case it is useful to consider nonconservative
systems, in particular systems with time-dependent external forces.
For such systems the physical structure changes with time, and so does
the mathematical structure that describes it. The observables, in particular
the Hamiltonian, will depend explicitly upon time; and thus the dynamical
laws will change with time. For such systems it is not possible to give the time
evolution by a unitary operator as in (1.15) and (1.16), or as in (1.22) and
(1.23). But one can still try to express it in a differential form in analogy to
(1.24) or (1.33). The Hamiltonian operator, which is no longer the energy
operator of the system, depends explicitly upon time. In this case the time
dependence may be formulated in the following form, which is suggested
from correspondence to classical mechanics:
dA(t) = ~ [A( ) H( )] oA(t) (1.59)

dt iii t, t + at '
where oA(t)lot is the derivative of A(t) with respect to its explicit time
dependence. If one assumes that every observable A is a function A =
A(Qi(t), Pj(t), t) of Qi(t) and Pit) with an additional explicit dependence
upon t, then (1.59) is obtained from the corresponding equation in non-
relativistic classical mechanics by replacing the Poisson bracket by 11ili
times the commutator. Equation (1.59) is a generalization of (1.24) and, in
the case of an isolated physical system, goes over to (1.24), which corresponds
to the classical expression for conservative systems. For systems that have
quantum-mechanical observables that cannot be expressed as functions of
Qi(t) and Pit), it is not clear whether a Hamiltonian operator can be found
such that the time development can be expressed by (1.59). Equation (1.59)
can no longer be integrated into a form analogous to (1.22); even for the
case that A does not depend explicitly upon time [oA(t)/ot = 0], there does
not exist a general integration theory for the solution of
dA(t) 1
dt
ih [A(t), H(t)]. (1.60)
XII.A Mathematical Appendix: Definitions and Properties of

Operators that Depend upon a Parameter
Let 6 be a linear space in which a limiting process is defined; such a space

is called a linear topological space. For example, in the Hilbert space 6 = :Yf
a limiting process is defined in the following way: A sequence ¢l' ¢2' ¢3' ...
is said to converge to ¢ E Jf; denoted
,I., .f{ ,I.,
't'n ...... ,+"
iff
II¢n - ¢II ~ 0 as n ~ 00. (A.l)
The Hilbert space :Yf is not the only linear topological space; if one has a
linear space, one can equip it with all kinds of topologies, i.e., one can give
all kinds of meanings to the convergence of a sequence. For example, instead
of one norm-as in the Hilbert space-one can define in a linear space 6 a
countable sequence of norms I I p'
11¢ll o ::; 11¢111::; 11¢11 2 ::; ... forall ¢E6,
and define convergence in the following way: A sequence of elements {¢n}
converges to ¢ with respect to the topology given by the countable sequence
of norms, denoted
iff for every p = 0, 1,2, ...

II¢n-¢llp~O as n~oo. (A.2)
Such a linear topological space 6 is called a countably normed space. In
quantum mechanics one usually uses the definition of convergence (A.l),
i.e., one uses the Hilbert space rather than any other linear topological space.
No physical justification for this choice (made by J. von Neumann) exists.
The notions of "continuity," "differentiability," etc. are defined with
respect to a certain topology or meaning of convergence. Hence continuity
in a Hilbert space has a different meaning from continuity in a countably
normed space or continuity in any space with another kind of topology. For
a given linear topological space 6, then, we say that an operator function
A(t) (i.e., an operator which depends upon a parameter t) is continuous at
t = to iff for any ¢ E 6
6
A(t)¢ ~ A(to)¢ as t ~ to' (A.3)
XII.A Mathematical Appendix 325
A(t) is said to be continuous iff (A.3) holds for all values of to' If one assumes
that in quantum mechanics the space of physical states is the Hilbert space,
then the vague statement made about the continuity of the time-evolution
operator U(t) has the precise meaning given by (A.3), where
"(3,, ",1f"
--+ means --+,
i.e., U(t) has the property
II U(t)4J - U(to)4J11 --+ 0 as t --+ to' (AA)
As mentioned in several places, in particular in Section 11.8, there is no

physical reason for which the Hilbert-space convergence is preferred over
any other definition of convergence, and therefore there is also no reason to
give the vague physical meaning of continuity the precise mathematical
definition (AA). One could as well have chosen any of the other topologies,
i.e., used the definition (A.3) with any other mathematical definition of con-
vergence in $. In the numerous derivations of the Hilbert space for quantum
mechanics, it is usually at this point (i.e., the choice of (A.4) for the vague
physical meaning of continuity) that the Hilbert space is sneaked in.
An operator function A(t) is called differentiable at t = to (with respect to
the topology on $) in a subspace ~ s;;; $ if for any 4J E ~
A(to + ~t) - A(to) ,J, S ./,
'I'--+'I'E,-,
c:
as
A
ut--+,
0
(A.5)
~t
i.e., if the limit of«A(to + ~t) - A(to»/~t)4J exists in $. The vector", defines
a linear operator which we shall denote by dA(t)/dt 11= 10'
'" = dA(t) I 4J,

dt 1=/0
and which is called the derivative (with respect to the topology on $) of
A(t) at t = to' Thus the derivative dA(t)/dt 1/=/0 is defined by
dA(t) I 4J = lim(S) A(to + ~t) - A(to) 4J. (A. 6)

dt 1=/0 41 .... 0 ~t
If one chooses the Hilbert-space convergence, then one writes
dA(t) I 4J = lim(Jf') A(to + M) - A(to) 4J. (A.7)

dt 1=/0 41 .... 0 ~t
The action of the derivative dA(t)/dtl/=/o is defined on all vectors 4J E ~ s;;; $,
i.e., on all vectors for which the limit (A.5) exists.
An operator function A(t) that satisfies
(A.8)
and is continuous with respect to the Hilbert-space topology is called a
one-parameter group of operators with respect to the Hilbert-space topology.
For such a one-parameter group of operators the following properties can

be proved:
1. A(t) is differentiable on a "dense" subspace 3 £1) ~ Yf. This justifies
(1.9), not for all vectors, but for all vectors in £1).
2. On £1), A(t) can be differentiated an arbitrary number of times.
3. The sequence {cPn}, where
t d A(t) I
k k
cPn = L kfd[k
k=O
n
• t=O
cP,
converges (with respect to the Hilbert-space topology) for all elements
cP of a dense subspace !Jj ~ ~ ~ Yf. This justifies (1.10).
Problems
1. H, a, and at are the operators as defined by (11.3.1) and (II.2.1b). Using the axiom
of time development (V), calculate da/dt and da t /dt.
2. The Hamiltonian H for a harmonic oscillator subject to an external force K(t)
is given by
p2 mro2
H = 2m + -2- Q2 - QK(t) = Ho - QK(t).
Here Ho, not the Hamiltonian H, is the total energy.

(a) Express H in terms of the operators a, at and the identity operator.
(b) Using the axiom on time development, calculate da/dt.
(c) Solve the differential equation obtained in part (b) and find the explicit ex-
pression for aCt).
3. For the harmonic oscillator under the external force K(t) in Problem 2, the
orthonormal eigenvectors !/tn of N = ata can be constructed in a manner analogous
to that used in the force-free case. One finds
1
!/tn(t) = c. [at(t)]"!/to(t),
v' n!
where !/to(t) represents the ground state. At the time t = 0, the energy of the harmonic
oscillator is measured to be llro/2. Denote the ground state at t = by 4>0 = !/to(t = 0).
(a) What is the statistical operator for the harmonic oscillator after the energy
°
measurement?
(b) Show that w. = I(!/t., 4> oW is the probability for measuring a value of the
energy E = llro(n + t) at some later time t.
(c) Show
I F(t) 12 • -IF(l )12

W. = -n-'-.-e ,
3 A subspace §Z is called dense if it differs from the space :If only by limit elements. Whereas
every infinite sequence with the property tP m- tPn --+ 0 for n, m --+ ::IJ (Cauchy sequence) has a
limit element tP in :If: tPn --+ tP E:If, (:If is complete); such a tP will in general not exist in §Z.
Problems 327
where
F(t) = ~ . S' K(r)e'"" dr.

I .
y 2mhw 0
(d) Show that the expectation value for the position operator immediately after the
energy measurement at t = 0 is
<Q> = ~I S' K(r) sin (t - r) dr.

mw 0
4. The harmonic oscillator under the influence of an external force

K(t) if t > 0,
K(t ) ={
o if t < 0
is at the time t = 0 in a state in which the energy is hwl2. The property that the energy
at a time t is equal to or smaller than hw(m + }) is characterized by the operator
A= 2::=0 A"'n(ri' where A"'n(r) are the projection operators on the one-dimensional
subspaces spanned by the t/ln(t).
(a) What is the probability for finding the energy to be equal to or smaller than
hw(m + !)?
(b) What is the statistical operator W' after the measurement of A with a positive
result?
(c) Show that
=0 for n' > n,

={ 1 m
= 2::=0 lin! IF(t)1 2n ' n~o It/ln(t»ei(n'-n)wr
1
x - 1 . - [F(t)]"[F*(t)]"" for n' ~ n.
y n!n'!
CHAPTER XIII
Some Fundamental Properties of

Quantum Mechanics
This chapter illustrates some of the characteristic features by which quantum

mechanics differs from classical theories. In Section XIII.1, using a gedanken
experiment with the Stern-Gerlach apparatus it is shown that a polarized
beam (a pure state) cannot be split by the magnetic field, and that the splitting
of such a beam is a consequence of the measurement. In Section XIII.2
we derive the predictions of quantum mechanics for spin correlation
measurements in the singlet state of pairs of spin-t particles. In Section XIII.3
these predictions are confronted with Bell's inequalities, which follow from
a hypothesis first proposed by Einstein, Podolsky, and Rosen. We conclude
with a short discussion of hidden-variables theories.
XIII.1 Change of the State by the Dynamical Law and by the

Measuring Process-The Stern -Gerlach Experiment
In the preceding chapter we have formulated how the state of a system
changes in time or how an observable changes in time as a consequence
of the dynamical law. This change is totally deterministic. This means that
once the observable (or the state) is known at the particular time, it can be
exactly predicted for all other times. In Section 11.4 we have described
another kind of change in time, the change of the state by a measurement.
These measurement processes are not deterministic and for their outcome
one can only make probabilistic predictions. Thus there are two completely
different processes of time change. In this chapter we shall illustrate these
328
XIII.! Change of the State- The Stern-Gerlach Experiment 329
Magnet I Magnet 2 Magnet 3
, U: U U
L ---- - B : ---- --~~
-:...-
A
-.. ---
Y - -- - -==-.:: -__ : - ......
___----- - -- ..... D
n in n
B+ I - - ___ - -
I
B
Figure 1.1 Stern-Gerlach experiment. Magnet I is used in a simple beam splitting
experiment; Magnet 2 and Magnet 3 are added in a gedanken experiment to reverse
the splitting of the beam.
processes by considering in detail a gedanken experiment with the Stern-

Gerlach apparatus, which was briefly described in Section IX. 1.
In the Stern-Gerlach experiment (Figure Ll), a beam of hydrogen atoms
in the ground state passes through a strongly inhomogeneous magnetic
field B. Under the conditions of this experiment the hydrogen atoms may
be considered to be the combination of two elementary physical systems
(cf. the basic postulate IVa of Section III.S): Physical system I is the ele-
mentary rotator with angular momentum i, and describes the spinning
electron; physical system II is the elementary particle that describes the
motion of the structureless hydrogen atom in the experimental setup.
The space of physical states of system I is the two-dimensional spin space
YlJ = its = fJlll/2 (cf. Section 111.3). As a basis for YlJ we choose the eigen-
vectors 1+ ) and 1- ) of the spin component S 3 in the direction of the magnetic
field B, with 1
S31+) = + il+) and S31-) = -il-)· (Ll)
The magnetic moment of this system is [Equation (lX.3.1S)]
e
M = -2--S = -2/1BS. (1.2)
2mec
The energy operator of system I is a constant; because of our freedom of
choice in defining the zero of energy, we may set this constant equal to zero:
Hl = O.
The space of physical states of system II is the space YlJ, spanned by the
generalized eigenvectors p) of the momentum operator P or by the general-
1
ized eigenvectors 1 x) of the position operator Q; Q and P are the position
1 We are again using units where h == I.

330 XIII Some Fundamental Properties of Quantum Mechanics
and momentum of the hydrogen atom considered as a whole. The energy

operator for system II is just the kinetic-energy operator
(1.3)
where M is the mass of the hydrogen atom.
System I and system II are coupled by the external magnetic field B;
according to Equations (IX 3. 1) and (lX3.18) the interaction energy operator
that represents this coupling is
Hint = -M· B = -(M ® J). (I ® B) = 2JiBS3B3(Q) = 2JiBS3B3(Q3)·
(1.4)
In (1.4) we have approximated the magnetic field B as having only a z-
component and as being a function of z = X3 alone? Consequently the energy
operator of the physical combination of these two systems, i.e., of the hydrogen
atom in the ground state with electron spin in the magnetic field, is
1
H = H, + HII + Hint = 2M p2 + 2Ji BS • B(Q)
In choosing this combination of I and II for the description of the hydrogen

beam in the magnetic field we have neglected (1) the magnetic moment of
the proton, (2) the influence of the magnetic field B upon the electron-
proton interaction in the hydrogen atom, and (3) the difference between the
electron position and the position of the hydrogen atom as a whole. The
justification of (1) and (2) may be seen by an examination of the orders of
magnitude involved: (1) The magnetic moment of the proton is [cf.
(IX3.18p)] Mp = (2 + 3.59)(e/2mpc)S= 5.59(me/mp)JiBS. Since the mass of
the proton is much greater than that ofthe electron (mp = 1836me), it follows
that the magnetic moment of the proton is almost three orders of magnitude
smaller than the magnetic moment of the electron; the interaction energy of
the proton magnetic moment in the magnetic field is therefore negligible when
compared to (1.4). (2) The Bohr magneton JiB = 5.795 X 10- 9 eV/gauss is
sufficiently small that even in a fairly strong field of 10 3 gauss the magnetic
interaction energy JiBB is six orders of magnitude smaller than the difference
between the lower energy levels of the hydrogen atom. Thus the internal
states of the hydrogen atoms do not change significantly and the operator
for the internal energy of the hydrogen atom [essentially the energy operator
(VI.3.2) of the Kepler problem] may be considered a constant for the present
problem, which we set equal to zero. Instead of the above described space
Yli" the space of states of system II should be £';, = Yli, ® f!ll, where f!ll is
the space of states for the Kepler problem [cf. (V1.5.2)]. But as we are keeping
the hydrogen atom in the ground state [i.e., in the one-dimensional subspace
2 The Maxwell equation V • B = oB;/ox; = 0 tells us that the variation of B\ with x is just
as great as the variation of B3 with z. We shall return to the subject of this approximation in the
appendix to this section.
XIII.1 Change of the State - The Stern-Gerlach Experiment 331
~(n = 1) c ~], the space of states for system II is in fact YCJ, @ ~ (n = 1),
i.e., is YCJ,.
°
The experimental arrangement is such that at t = a pulse of hydrogen
atoms enters the region of the magnetic field at the point A (cf. Figure 1.1)
and moves with an average momentum p in the y direction. Idealizing, we
assume that in this experimental arrangement a pure state of system II has
been prepared. This state is described by the state vector (statistical operator)
(1.6)
in the Heisenberg picture for all times t z 0, or by
(1.7)
in the Schrodinger picture at time t = 0. System I may be in a pure state with
statistical operator
WI = 14»<4>1 where 14» = al+ >+ [31- > [3 E C) (a, (1.8)
if the beam has been polarized before it reaches point A. (The state 14»will
be an eigenstate of fi . S for some direction fi determined by the choice of
a and [3.) Alternately the system I state may be a mixture, e.g.,
1tI = 1(1 + >< + I + I- >< - I) = y, (1.9)
where I here is the identity operator in the spin space YCJ = -is. The state of
the combined physical system is described by
w= W, @ W" or Ix> = 14» @ ItfJo>, (1.10)
the second description in terms of the state vector Ix> being an
alternative
to the description in terms of the statistical operator W in the case that
system I is in a pure state. 3
In considering the time development of the pulse of hydrogen atoms we
shall use both the Heisenberg and the Schrodinger pictures. By use of
(XIL1.59), (1.5), and the canonical commutation relations, we see that the
Heisenberg equations of motion for the position and momentum operators
are
dQ(O 1 1
----;{t = i [Q(t), H(t)] = M pet), (1.11)
dP(t) 1 · 1 (8B
---;{t = i [pet), H(t)] = i 2,uB S 3[P(t), B 3(Q3(t»] = 0,0, -2,uB S 3 8Q3(t) .
3 )
(1.12)
3 Remark on the state of a combination of physical systems: Recall that the operators in the
direct-product space are given by Equation (IlI.S.7). A general statistical operator in .#I @ .#II
is therefore not given simply by the direct product of statistical operators in .#I and in .#II.
However, if either of the states Wi or Wil is a pure state, then Wi and Wil uniquely determine the
state W of the combined system as the product W = Wi @ Wil. This is the situation in the prob-
lem under investigation. W is also given by W = Wi @ Wil if the state of the combined system
is determined by measurements that have been performed upon systems I and II separately. If,
however, measurements are performed that measure correlated properties of systems 1 and 11,
then in general W is not factorable into the form WI @ Wil.
[The meaning of a/aQ; was given in the paragraph following Equation

(XII.1.29).] Let us first investigate the case that system I is in the spin state
> >,
1+ or 1- so that the state of the combined system is
Ix±> == I±> ® 1"'0>· (1.13)
The motion of the expectation value of the momentum operator in these

states is then given by
(1.14)
Thus if the state has spin up, S3 = +t (spin down, S3 = - t), i.e., if the state
>
is 1+ (1- »,
then the time development of the expectation value of the
momentum's z-component fulfills
d _ aB 3
dt <X ± I P 3(t)lx ± >= +JJB<"'ol aQ3(t) 1"'0>
- J
= +JJB d 3 x aB
3
aX 3 <"'olxt><xtl"'o), (l.15±)
where the 1x t> are the time-dependent eigenvectors of Q in the Heisenberg

picture. [The upper (lower) sign in (1.15 ±) refers to the spin-up (spin-down)
case.] The right-hand side of(1.15 ±) shows that the change of the momentum
depends upon the particular properties of the magnetic field; let us assume
for simplicity that
aB 3
- = const > 0. (1.16)
aX 3
Then
(1.17±)
where
(1.18)
Using the initial condition that the particle has momentum

<X ± 1 pet = 0) 1X ± >= (0, p, 0) at t = 0,
we may integrate (1.17 ±); the expectation value of the momentum operator
at time t is then
<X ± IP(t)lx ± >= (0, p, +Kt). (1.19±)
XIII.! Change of the State-The Stern-Gerlach Experiment 333
If we take the expectation value between I X ±) of (1.11) and use the initial
condition <X ± IQ(t = 0) IX ±) = 0 together with (1.19 ±), then the
expectation value of the position operator is
1
<X ± IQ(t) IX ±) = M (0, pt, =+= Kt 2/2). (1.20±)
Thus the expectation value of the position operator, i.e., the position of the
pulse of hydrogen atoms, moves on a parabolic orbit that bends downward
(upward) in the case of a pure spin-up (spin-down) initial state, as illustrated
in Figure 1.1. [This presumes that K > 0, i.e., presumes (1.16).]
Let us now describe the same situation in the Schrodinger picture, where
the state changes in time. At the time t = 0 the state is given by
Ix ± (t = 0» = Ix ±) = I ±) ® It/!o), (1.21 ±)
which describes the pulse of hydrogen atoms with spin up (spin down) at
the position A. According to (XII. 1.35), the time development of this state is
I X ± (t» = Ut(t)(1 ±) ® It/!o» = e- it (Hn+2I'B S 3B3)(1 ±) ® It/!o»

= e- it (HIdI'BB 3)(1 ±) ® It/!o» = I ±) ® e- it (HIdI'BB3) It/!o)·
(1.22±)
Thus the initial pure state with spin up will move through the magnetic field
according to
Ix + (t» = 1+) ® e- ii (Hn+I' BB3)It/!o) = 1+) (8) It/! + (t», (1.23+)
in contrast to the motion
of the pure state with spin down. In (1.23±) we have defined
(1.24±)
Now the relations connecting the momentum operator P, the position

operator Q, and the state vectors Ix ± (t» of the SchrOdinger picture with
pet), Q(t), and IX ±) of the Heisenberg picture are
P = pet = 0) = Ut(t)P(t)U(t), (1.25a)
Q = Q(t = 0) = ut(t)Q(t)U(t), (1.25b)
Ix ± (t» = Ut(t)lx ± (t = 0» = Ut(t)lx ±), (1.25c)
where U(t) = eitH = exp(it(Hn + 2J1BS3B3(Q3»)' We may write the

Schrodinger-picture analogue to (1.14) by taking the expectation value of
(1.12) between the state vectors X ±) and shifting the time dependence
1
from the observables to the state vectors:

d d
dt <X ± (t)I P 3Ix ± (t» = dt <X ± IP3(t)lx ±)
dP 3 (t) aB 3
= <X ± 1~ 1X ±) = <X ± 1 - 2,uB S 3 aQ3(t) 1X ±)
= - 2,uB< ± 1S31 ± )<1/10 1eit (Hu±I'B B3) ~~: e- it(HtdI'BB3) 11/10)

aB 3
= +,uB<1/I ± (t)1 aQ3 11/1 ± (t».
°
(Note that [S3, H] = for the Hamiltonian H of (l.5); the matrix element
<± 1S31 ±) thus has no time dependence.) If we again assume (1.16), then
<1/1 ± (t)1 aB
aQ: 11/1 ± (t» = J d3x aB
ax: <1/1 ± (t)lx)<xll/l ± (t»
= aaB3 <1/1 ± (t) 1/1 ± (t» = aaB3 .

1
X3 X3
Consequently
d
dt<x±(t)lPlx±(t»= (0,0,+,uB
a B 3 ) =(O,O,+K). ( 1.26)
aX3
The expectation value
:t <X ± IQ(t)lx ±) = ~ <x ± IP(t)lx ±)

of (1.11) easily transforms to the Schrodinger-picture analogue
d 1
dt <X ± (t)IQlx ± (t» = M <X ± (t)lPlx ± (t». (1.27)
Equations (1.26) and (1.27) together with the initial conditions
<X ± (t = O)lPlx ± (t = 0» = (0, p, 0)

and
<X ± (t = O)IQlx ± (t = 0» = (0,0,0)

lead to the Schrodinger-picture analogue of (1.20 ±):
x±(t) == <X ± (t)IQlx ± (t» = M1 (

O,pt, + 2Kt2) . (1.28 ±)
As we would expect, the result is the same as in the Heisenberg picture.

The probability distribution for the position is the "expectation value"
of the operator I ® I x) <xl. More precisely, the probability of obtaining a
value in the volume (x - €, X + €) in a position measurement is the
expectation value of the projection operator
10J\(x,€)=I® f x +£
x-£ d 3 x'lx') <x'i (1.29)
[cf. Equation (11.8.8)]. Thus the position probability distribution in the

state Ix + (t» = 1+) ® 11/1 + (t» is given by
w+(x, t) = < + 1+ )<1/1 + (t)lx) <xl 1/1 + (t» = <1/1 + (t)lx) <xll/l + (t).
(1.30+ )
In particular, the probability of obtaining a value in the volume (x - €, X + €)
at time t is
W +(x - €, X + €, t) = f
X
x-£
+£
d3 x' <1/1 + (t) Ix') <x' I1/1 + (t). (1.31 +)
The probability distribution moves with time. The exact motion of w +(x, t)
may be obtained by solving the SchrOdinger equation for <x I 1/1 + (t».
From our considerations concerning the expectation value (1.28 + ), however,
we have a rough idea of how w+(x, t) moves: At t = 0, w+(x - €, X + €, t)
and W +(x, t) will be essentially zero everywhere except in a small volume
around the point A = (0, 0, 0). From A the pulse moves along the lower
parabola, given by (1.28 + ); at the time tB = MYB/P (i.e., at the time tB at
which the pulse has the y-coordinate YB), w+(x, t) will be essentially zero
everywhere except in a small volume around the point B+ = (0, ptB/M,
-Kt~/2M).
Similarly, the probability distribution
W_(x,t) = <1/1- (t)lx)<xll/l- (t) (1.30- )
r::
and the probability
w_(x - €, X + €, t) = d 3 x'<I/1- (t)lx')<x'll/I - (t) (1.31- )
of finding the system in the volume (x - €, X + €) are the ones for the state
Ix - (t» = 1-) ® 11/1- (t), which was a pure state with spin down at A.
Then w_(x, t) and w_(x - €, X + €, t) move along the upper parabola
(1.28 - ), and will be significantly nonzero at time tB only around the point
B_ = (0, ptB/M, + Kd/2M).
These facts enable us to perform a measurement of the polarization ~a

measurement of the system-I observable S 3 ~ by measuring the system-II
observable Q. If we put a detector plate at B (at a distance YB = ptB/M
from A), then it will record the presence of hydrogen atoms at B + =
(0, ptB/M, -Kd/2M) or at B_ = (0, ptB/M, +Kt~2M) only if one started at
A = (0,0,0) with the pure spin-up state IX + (t = 0» = 1+> ® 11/10> or
with the pure spin-down state Ix - (t = 0» = 1- > ® 11/10>, respectively.
Thus a measurement of Q3 on system II with result z = z +(t B) = - Kt~/2M
constitutes a measurement of S 3 on system I with result S3 = +! and a
measurement of Q3 on system II with result z = L(t B) = +Kt~/2M
constitutes a measurement of S 3 on system I with result S3 = -l
We now investigate the case where the system-I state is the mixture (1.9),
where the coefficients! before each of the projectors I + > < + I and I - > < - I
represent the equal a priori probabilities of the system being in the states
1+> or I - >. At t = 0 the beam is described by the statistical operator (UO)
with Jt(t = 0) given by (1.9) and Jt([(t = 0) given by (1.7):
W(O)= Jt(0) ® Jt([(0) = (!I + >< + 1+ !I- >< -I) ® (11/10><1/101)
= !I + > ® 11/10>< + I ® <1/10 I + !I- > ® 11/10>< - I ® <1/10 I
=!I X+ (O»<X + (0)1 +!I X - (O»<X - (0)1. (1.32)
The time development of W is
Wet) =!I + > ® 11/1 + (t» < + I ® <1/1 + (t)1
+ !I- > ® 11/1 - (t»< -I ® <1/1 - (t)1
= !Ix + (t»<x + (t)1 + !Ix - (t»<x - (t)l· (1.33)
The probability distribution w(x, t) = Tr«(I ® Ix><x I)W(t» for the position
is then
w(x, t) = hxll/I + (t»<1/1 + (t)lx> + !(xll/I- (t»<I/1- (t)lx> (1.34)
= !w+(x, t) + !w_(x, t), (1.35)
where w+(x, t) and w_ex, t) are the probability distributions of (1.30±),
and their (significantly) nonzero parts move downward and upward along
the parabolic trajectories (1.28 + ) and (1.28 -), respectively. Thus the un-
polarized beam is split into two beams.
This was in fact what Stern and Gerlach observed, in contradiction to the
predictions of classical physics. 4 Classically an unpolarized beam contains
atoms with magnetic moments equally distributed in all directions. The
deviation would be proportional to the magnitude of the z-component of
the magnetic moment, and hence one would expect the narrow beam to
just smear out in the z-direction.
We can now add (at least in a gedanken experiment) second and third
magnets, or another suitable arrangement of magnets, that reverse the
splitting effect of the first magnet. The two beams are then brought together
" The original Stern-Gerlach experiment (1922) used silver atoms rather than hydrogen
atoms.
XIII.! Change of the State-- The Stern-Gerlach Experiment 337
again; and at D at some later time t D we again have the same state as at
A at time t = 0, i.e., a mixture described by (1.32). If this pulse is now passed
through another Stern-Gerlach apparatus with the magnetic field in the
x-direction, it will split into two beams. This is easily seen by representing
"'I as
~ = }! = ~ISI = +~)(Sl = +~I + ~I SI = -~>(SI = -~I, (1.36)
where lSI = + D and lSI = - ~) are the basis vectors of Jt( satisfying
S11 s 1 = ±~) = ±~ISI = ±~); (1.37)
the above arguments are then repeated with the z-component exchanged
with the x-component.
We will now analyze the Stern-Gerlach experiment for the case that the
state at A is a pure state, polarized in any direction n other than the z-direction,
e.g., the x-direction. 5 At time t = 0 the state is then described by the statistical
operator
W(O) = Ix(O» (x(O) I = 14>)(4)1 ® It/Jo) (t/Jo I (1.38)
or by the vector
Ix(O» = 14» ® It/Jo) = 0:1 +) ® It/Jo) + fil-) ® 11// 0 ), (1.38')
where 14» is given by (1.8). This state develops in time according to (XIl.l.35);
using the above results (1.22 ± ), (1.23 ± ), and (1.24 ±) one obtains
Ix(t» = 0:1 +) ® It/J + (/» + fil-) ® It/J - (t». (1.39)
Now in a unitary time development a pure state goes into a pure state. The
system will therefore remain in a pure state while it passes through the mag-
netic field; there is no splitting into two subensembles. If the beam reaches
D after having passed through the reverse splitting magnets, the pulse is
again in the pure state with the same polarization that it had at A. For
example, if we start at A with a beam polarized in the x-direction, it will
again be polarized in the x-direction at D at the time t D ; if it is then sent
through a subsequent Stern-Gerlach apparatus with the field in the
x-direction, it will not split, but will just bend.
Suppose now that we perform a position measurement on this pure state at
B, i.e., we place at B a measuring device (screen) that lets the beam pass
through but that registers the presence of hydrogen atoms. Recall that the
probability of finding the hydrogen atom in the volume (x - €, X + €)
around x is given by the expectation value of the operator! ® A(x, €) of(1.29)
with the statistical operator
Wet) = Ix(t» (x(t)1 = 10:1 21+) ® It/J + (t»( + I ® (t/J + (t)1
+ IfiI21-) ® It/J - (t»( -I ® (t/J - (t)1
+ o:PI +) ® It/J + (t»( -I ® (t/J - (t)1
+ fifJl-) ® It/J - (t»( + I ® (t/J + (t)1 (l.40)
5 Use of the Pauli spin matrices easily gives Is, = ±!> = (llfi)(1 +) ± 1-») (up to
an arbitrary phase factor). For the Pauli spin matrices see Problem 111.5.
corresponding to the pure state (1.39); thus
wlx(t»(x - €, X + €) = Tr(W(t)(J ® A(x, €)))

= lal 2 Tr,(1 +)< + I) Tr,,(11/i + (t)<1/1 + (t)IA(x, f))
+ I/W Tr,(I-)< -I)Tr,,(II/I- (t)<I/1- (t)IA(x,€))
= laI 2 w+(x - €, X + €, t) + 1131 2 w_(x - €, X + €, t),
(1.41 )
where w±(x - €, X + €, t) is given by (1.31±). The probability distribution

for the position, i.e., the "expectation value" of the" operator" I ® Ix)<x I,
is obtained in the same manner, and is given by
(1.42)
For the case that the hydrogen atoms were originally polarized in the
x-direction (i.e., lal 2 = 1 and 1131 2 = 1), we have
wlx(t»(x) = 1w+(x, t) + 1w_(x, t).

Thus the position probability distribution for the pure state (1.8) is the
same as the probability for a position measurement in the mixture with the
system-I statistical operator
(1.43)
From our discussion of the motions of w +(x, t) and w _(x, t) we conclude
that the hydrogen atoms pass through the screen at B + and at B _ with
relative probabilities Ia 12 and 1131 2, respectively. Now Equations (1.40),
(1.41), and (1.42) depend upon the assumption that the system is in the pure
state Ix(t) given by (1.39). This assumption is valid only for t < t B, for after
the position has been measured at B the state is no longer given by Ix(t);
it is changed to a new state by the measurement process. If the screen has
openings only around B + and B _, then, according to the basic postulate
IIIb [Equation (l1.4.51)J, the state is given at time t = tB by
WB(tB) = A(xB+, €)lx(t B) <x(tB)IA(x B+, €)

+ A(XB-, €)lx(t B)<x(tB)IA(xB- ,€). (1.44)
Now
= f
XB±
XB±-€
+€
d 3 x' d 3 x"lx')<x' I1/1 ± (tB)<1/I ± (tB)lx")<x"l
Since <x'll/I- (t B) = 0 for x' near x B+ and since <1/1- (tB)lx") = 0 for
x" near XB+, it follows that
Similarly
Expansion of (1.44) by means of (1.39) and use of the last three equations
allows us to rewrite (1.44) as
WB(tB) = Ial 2 1+ ><+ I ® A(XB+, £) Il/I + (tB» <l/I + (t B) IA(xB +, £)
+ IfJ 12 1_ ><-I ® A(XB_, £) Il/I - (tB» <l/I - (tB) IA(x B _, E).
(1.45)
We thus see that the performance of a measurement changes the situation

drastically. If the beam in the pure state (1.39) passes B without a measure-
ment being performed, then the system remains in this pure state Ix(t». If a
measurement is performed as the beam passes B, i.e., if it is registered that the
hydrogen atom has passed through the opening at B + or at B _, then the
system becomes a mixture of the electron spin-up and spin-down states. For
example, if the openings in the screen are symmetrical with respect to the xy
plane,
W+(X B+ - £, XB+ + £, t B) = W_(XB_ - £, XB- + £, t B), (1.46)

and are big enough to let all of the particles in the beam pass through, then
The state after the measurement then becomes
WB(t) = lal 2 1+ >< + I ® Il/I + (t»<l/I + (t)1

+ IfJ1 2 1- >< -I ® Il/I- (t»<l/I- (t)1
= lal 2 1+ > ® Il/I + (t»< + I ® <l/I + (t)1
+ IfJ1 2 1- > ® Il/I- (t»< -I ® <l/I- (t)1
= lal 2 1x + (t»<x + (t)1 + IfJI 2 1x - (t» <x - (t)1 = WM(t)
for t > tB • (1.48)
If the state (1.48) now passes through the reverse splitting magnets, it will
arrive at D in a mixture of spin-up with spin-down states. A subsequent
Stern-Gerlach apparatus with magnetic field in the x-direction will then
split the beam.
In summary, we have found that the pure state (1.38) is not split by the
inhomogeneous magnetic fields of the first and second magnets unless a
position measurement is made at the plane B. If such a measurement is
made, however, the pure state splits at B into two subensembles of spin up and
spin down, and consequently into two subensembles of spin right (s 1 = + t)
and spin left (SI = -t), by a subsequent Stern-Gerlach apparatus, whose
field is in the x-direction. This concrete example illustrates the content of
the basic postulate IIIb.
Appendix to Section XIII.1

In (1.4) and (1.12) we have neglected the force upon the particle except for
the force in the z-direction. This requires a justification, because due to
the Maxwell equation, V' B = 0, a large oB 3/oX 3 will also require a large
oBI oB 2 oB 3
-+-=-- (l.49)
oX I oX 2 OX3'
and therewith a large force perpendicular to the z-direction. We shall describe
in the following the simplest theoretical arrangement in which the
time average of the force on the particle perpendicular to the z-direction
never becomes large. This can be achieved if one chooses a large magnetic
field in the z-direction. This field causes a Larmor precession of the magnetic
moment, as a consequence of which the time average of the force perpen-
dicular to the z-direction is small-not the force itself.
We simplify matters by assuming the magnetic field does not vary with X 2
(no edge effects); then (1.49) reduces to
(1.50)
and the Maxwell equation V x B °

= reduces to
oB 3
-;-- = 0, (1.51 )
uX 2
and
oB 3 oBI
(1.52)
OXI OX3·
A magnetic field compatible with this assumption and which also leads to
(1.16) must be of the form
(1.53)
°
where b > and B2 are constants and I/J(X3) and ¢(x l ) are arbitrary functions
except that they must satisfy
ol/J(X 3) o¢(x l )
(1.54)
OX3 OXI'
which follows from (1.52). We choose B2 = 0, ¢(x 1 ) = 0, and l/J(x 1 ) = Bo

with
(l.55)
Therewith we have a magnetic field which has a strong component in the
z-direction and is inhomogeneous in the x- and the z-directions:
(1.56)
XIII.1 Change of the State~ The Stern-Gerlach Experiment 341
In an actual experimental set-up the field will, of course, be more com-

plicated.
For the electron, spin and magnetic moment are connected by (1.2) and the
Hamiltonian is given by (1.5). Using these along with the Heisenberg
equation of motion (XII. 1.24), one obtains
dM· 1
dt' = ih (-2J18)(2J18)[Sj' SJB i • (1.57)
From this and from the commutation relations of the Sj' there follows
dM
Tt = -2J18 M x B. (1.58)
For a static magnetic field, (1.58) has the solution

M(t) = M(O) cos wt +Bx M(O) sin wt
+ B[B' M(O)J(l - cos wt), (1.59)
where B = BIB, B = ft, and w = 2J18B is the Larmor frequency. This
solution describes the precession of M around the direction of B; the com-
ponent of M along B does not change in time.
For the magnetic field (1.56) with (1.55), the solution of (1.58) is given
to good approximation by
M(t) = M(O) cos wt + e3 x M(O) sin wt
+ e3[e3 . M(0)](1 - cos wt) (1.60)
with the frequency
w = 2J1BBO' (1.61)
In component form, (1.60) reads
M I (t) = M I (O) cos wt - M 2 (0) sin wt,
MzCt) = M I (O) sin wt + M 2 (0) cos wt, (1.60')
M 3(t) = M 3(0),
which describes the precession of the magnetic moment operator M around
the z-axis e3 with the frequency (1.61); the component M 3 does not change
in time.
Instead of (1.4) one has now
Hint = -M I(t)bQI - M3 Bo + M 3bQ3' (1.62)
and instead of (1.12) one has for the force operator
d~~t) = _ f [pet), QIJM 1(t)b + ~ [pet), Q3JM 3b. (1.63)
The second term on the right-hand side is the same as the term in (1.12),
except that we have here the special case
dB3
dQ3(t) = b. (1.64)
As a consequence the expectation value of P 3 (t) is given by (1.19).

The first term in (1.63) results in a component of the force operator in the
x-direction:
which oscillates with the frequency (1.61). Though its expectation value may
be of the same order as (1.17), it oscillates rapidly under the condition
(1.55), and so does the expectation value of PI (t). The time average of the
force as well as the momentum in the x-direction is zero. In an experiment,
which always measures the average value ofthe expectation value over a finite
time interval, a deviation from zero in the x-direction will not be observed.
In the conventional Stern-Gerlach experiment 6 the magnetic field is not
given by the idealization (1.56) but is usually of the form
B = B1(X 1, x3)e 1 + (Bo + B 3 (x l , x 3»e3
with BI(x l , x 3 ) and B 3 (X I , x 3 ) fulfilling (1.50) and 0.52) and IBo I ~ BII ;:::;
IB21· Clearly, the same arguments as above apply, and (1.4) suffices to
describe the observable effects.
XIII.2 Spin Correlations in a Singlet State
In the previous section we discussed the use of the Stern-Gerlach device

for the measurement of spin components of a single spin-1 system. In
particular we have described how the spin component of the physical system I
(the spin degrees of freedom of an electron, an elementary rotator) can be
measured by a position measurement on the physical system II (the center-
of-mass motion of a hydrogen atom) to which it is attached. System II
is used as an auxiliary device which permits us to use the position measure-
ment on II as a measurement of the spin component of system I along the
field. We can, therefore, consider system II together with the Stern-Gerlach
device as an apparatus for the measurement of the spin component of
system I.
Let a denote the unit vector in the direction of the inhomogeneous
magnetic field. In Figure 1.1 a = e3, the z-direction, but one may choose
any direction orthogonal to the particle beam. Then the Stern-Gerlach
device measures the spin component of system I along a, i.e., the observable
a . S. The measured value is + 1 (- 1) if system II is deflected "down"
Cup") and recorded by a counter in position B + (B _), cf. Figure 1.1. It
is more convenient now to consider instead of a . S the observable
~ = 2a' S, (2.1)
(, For other arrangements of a Stern-Gerlach experiment, see Myer Bloom and Karl Erdman,
Can. J. Phys. 40, l79 (1962).
which has the eigenvalues ± 1 instead of ±l We denote the eigenvectors

of ¢ by la±), with
¢Ia±) = ±lla±). (2.2)
In particular, the eigenvectors 1+) and 1-) of S3 in (Ll) can be more
explicitly denoted by 1e3 + ) and 1e3 - ), so that (Ll) now reads
The vector a can be obtained from the vector e 3 by a rotation specified

by a rotation angle 8 and a unit vector () parallel to e 3 x a, which determines
the direction of the rotation axis. The vectors 1a ±) can then be obtained
-up to an arbitrary phase factor-from the vectors 1±) == 1e3 ±) by
the transformation which represents this rotation:
la±) = e- ie ·s 1±),
where 0 = 80, and where
Since (see Problem 1)

e s 8 8
e- I' = cos - - i20 . S sin -
A
2 2'
one has
( 8
la±)= COS2-i20·Ssin21±).
A 8)
Thus 1a ±) is just another basis system of the space ~s~ 1/2; whereas the
basis 1±) was adapted to the Stern-Gerlach apparatus with its inhomo-
geneous magnetic field in the e3 direction, 1a ±) is the basis adopted to a
Stern-Gerlach apparatus with an inhomogeneous magnetic field in the
direction of a.
Until now we have only discussed the measurement of spin components
of a one-particle system. We now consider the combination of two different
spin-! systems each described by ~s~ 1/ 2 • The state space for this combination
of the spin degrees of freedom of both particles (of two different elementary
rotators of spin!) is-according to the basic postulate IVa-~(~) 1/2 ®
~(;; 1/2: We will discuss the measurement of spin correlations on this system,
i.e., of the correlations between certain spin components of these two
particles. A system of basis vectors in this space is
1(1)+) ® 1(2)+), 1(1)-) ® 1(2)-),

(2.3)
1(1)+) ® 1(2)-), 1(1)-) ® 1(2)+),
where (1) and (2) refer to the first and second particle, respectively, and
+ or - specifies the z component of the spin as in (1.1). Instead of the basis
system (2.3) one can also take-in the notation of (2.2)-the basis system
1(1)a+) ® 1(2)b+), 1(I)a-) ® 1(2)b-),
1(1)a+) ® 1(2)b-), 1(1)a-) @ 1(2)b+),
(2.3')
with two arbitrarily chosen but fixed unit vectors a and b.
Let Sla) denote the spin operators of the particle (:x) acting in ~(a~ 1/2,
:x = 1 or 2. The observable "2 x spin component of particle (1) along
a" is then given by
(2.4)
and the observable" 2 x spin component of particle (2) along b" is given by
(2.5)
where 1(') is the unit operator in ~(a~ 1/2. The observables (2.4) and (2.5)
commute, 7 and the basis vectors (2.3') are simultaneous eigenvectors of
both of them, with eigenvalues + 1 or -1. Because the observables (2.4)
and (2.5) commute, they can be measured simultaneously. (We will discuss
below how such simultaneous measurements can actually be performed.)
The possible measured values for each of them are + 1 or -1; the measure-
ment of d(1) (8) 1(2) in the four states (2.3'), for instance, will always yield
+ 1, -1, + 1, and -1, and the simultaneous measurement of 1(1) @ ~(2)
in the same states will yield + 1, -1, - 1, and + 1, respectively.8 For such
simultaneous spin measurements we may define, in addition, a spin correlation
observable. The measured value of this observable in a "single" measurement
of both ¢(1) ® 1(2) and 1(1) ® ~(2) is, by definition, the product of the value
obtained for ¢(1) @ 1(2) and the value obtained for 1(1) ® ~(2), and is there-
fore either + lor-I. This spin correlation observable is described mathe-
matically by the operator
¢(1) ® ~(2) = 2a . 8(1) @ 2b· 8(2). (2.6)
The basis vectors (2.3') are also eigenvectors of the operator (2.6); the
eigenvalues are + 1 for the first two and -1 for the last two basis vectors of
(2.3'). These eigenvalues coincide, according to the above definition, with
the measured values of the spin correlation observable in the four states (2.3').
Such simultaneous spin measurements on two-particle systems are
performed with two Stern-Gerlach devices. This is possible, however, only
if the two particles of each pair are spatially separated, and each particle
moves along a certain fixed axis, as shown in Figure 2.1. Only in this case
may two separate Stern-Gerlach devices be applied to the two particles
7 In fact, they are a complete system of commuting observables in 9l(I) 1'2 ® 9l({) 1/2.
H Remember that, according to Section 1I.4, the value obtained from a single measurement
of an observable on a system in any state is always one of the eigenvalues of the corresponding
operator. If the observable is measured on a system described by an eigenstate of that operator,
the measured value is always the corresponding particular eigenvalue.
B ___
;'::ffi ---(I) - -[@--m- ---0:: ~ ~

" a b .... .... B _
~ S '
B+_ B+
Figure 2.1 Simultaneous spin measurements on particle pairs (I) + (2). S is the particle
source, B ± are particle detectors, and a and b are field directions of the Stern-Gerlach
magnets.
(1) and (2) in such a way that the orbital motion (system II, in the terminology
of Section XIII.l) of each particle is used for a spin measurement on that
particle (as discussed in Section XIII.1). Imagine, therefore, a particle
source emitting pairs of particles, one pair at a time, such that particle (1)
is always emitted to the left, and particle (2) is always emitted to the right. 9
Then a Stern-Gerlach device with inhomogeneous magnetic field along
some direction a (perpendicular to the beam) may be applied to the left
beam, consisting of particles (1), and another Stern-Gerlach device with
field direction b may be applied to the right beam of particles (2), cf.
Figure 2.1. Each Stern-Gerlach device has two counters, one at a position
corresponding to B + in Figure 1.1 that detects all particles deflected against
the field direction a or b (" down "), and another one at B _ that detects all
particles deflected in the field direction (" up "). Then, for instance, a click
in the "up" (" down") counter of the left Stern-Gerlach device means that
the measured value of the" spin component" ¢(1) ® 1(2) of the single particle
(1) is -1 ( + 1).10 Similarly, the right Stern-Gerlach device detects the spin
component 10 ) ® ~(2) of the particles (2). For simultaneous measurements
of both ¢O) ® 1(2) and 1 0 ) ® ~(2) on single pairs, the four counters have to
be used in coincidence. Since the particles (1) and (2) are emitted pairwise
by the source, the two particles of a single pair pass the two Stern-Gerlach
magnets and arrive at two of the four counters (almost) simultaneously.
Therefore, e.g., a simultaneous response of the left "down" and the right
"up" counter means that a simultaneous measurement of ¢O) ® 1(2) with
the result + 1 and of 10 ) ® ~(2) with the result - 1 has been performed on
a single particle pair. The value of the spin correlation observable ¢(l) ® ~(2)
for this single pair is (+ 1)( -1) = -1. This kind of measurement is repeated
N times (N ~ 1), and the following numbers are recorded: the number
N + + of simultaneous clicks of the left "down" and the right "down"
counter, and the analogous numbers N + _ , N _ + , N __ counting the simul-
taneous clicks left "down" and right "up," left "up" and right "down,"
left" up" and right" up," respectively. The measured average values for the
9 We assume here that the particles are not identical. If the particles are identical one arrives
at the same results by a somewhat more complicated calculation.
[0 This is only true if the spin and the magnetic moment are antiparaJlel as for the electron;
otherwise the roles of the two counters must be interchanged. A factor 2 is included here in
the definition of "spin components."
observables ¢(l) ® [<2), 1(1) ® ~(2) and ¢(l) ® ~(2) in this run of N measure-
ments (here denoted by E1(a), EzCb) and E(a, b)) are then given by
1
E1(a) = N(N++ +N+_ -N_+ -N __ ), (2.7a)
1
Ez(b) = N(N++ -N+_ +N_+ -N __ ), (2.7b)
(2.7c)
of course, N = N + + + N + _ + N _ + + N __ .
According to quantum mechanics, these measured average values should
coincide with the expectation values of the corresponding observables in the
common spin state of the particle pairs emitted by the source. This spin
state depends on the nature of the source. Since ~1/2 ® ~l/Z = ~l EB ~o
(cf. Section V.2), the combination of two spin-~ systems may lead to total spin
1 or total spin 0. We will assume here that the particle pairs emitted by the
source have total spin zero, and are therefore in the singlet state
1
¢ = vf2(1(1)+ > ® 1(2)- > - 1(1)- > ® 1(2)+ », (2.8)
cf. Equation (XI.1.18). A source emitting particle pairs in this spin state ¢
might contain, for example, a large number of unstable compounds (decaying
states) of particles (1) and (2) with total spin zero, which are approximately
at rest. Every decay of such a compound produces two particles, (1) and (2),
which move in opposite directions away from the original position of the
compound; however, they are still in the common spin state (2.8). Using
slits and filters, then, one can select those particles (1) which move in a
fixed direction (left) and the accompanying particles (2) moving in the
opposite direction, in order to obtain a source of pairs with the desired
properties.
A straightforward calculation (Problems 2 and 3) now yields the expecta-
tion values of the three observables considered above in the particular
spin state (2.8). The result is:
<¢(l» == <¢I¢(l) ® /(Z)I¢> = 0, (2.9a)

<~(Z» == <¢I/(l) ® ~(Z)I¢> = 0, (2.9b)
<¢(l) ® ~(2» == <¢I¢(l) ® ~(Z)I¢> = -a· b. (2.9c)
These are-according to basic postulate II -the theoretical predictions of
quantum mechanics for the averages over many single measurements,
(2.7a)-(2.7c).
Comparing (2.9a) with (2.7a), we see that quantum mechanics predicts
that N + + + N + _ and N _ + + N __ (the numbers of cases in which the
spin of particle (1) was found to be parallel and antiparallel to a, respectively)
XIII.3 Bell's Inequalities, Hidden Variables, and the EPR Paradox 347
will be equal for any choice of a. Since the spin state </> is rotationally in-
variant, we could have anticipated this result. The same argument applies
to the comparison of (2.9b) and (2. 7b).
At first glance, the result (2.9c) does not surprise us either. Consider,
e.g., the particular choice a = b for the orientation of the two Stern -Gerlach
instruments; then (2.9c) becomes
<~(1) ® ¢P» = -1. (2.10)
Comparing this with (2.7c), and remembering that N = N + + + N +_
+ N _ + + N __ , we see that the quantum mechanical prediction E(a, a) = -1
implies N + + = N __ = 0; i.e., if the spin of particle (1) is found to be
parallel to a, the spin of the accompanying particle (2) of the same pair
cannot be parallel to a and is thus always found to be antiparallel to a, and
vice versa; or, more briefly: the spin components of the two particles along
an arbitrarily chosen direction a are always opposite to each other. This we
would also have intuitively expected: since each pair has total spin zero, the
spins of particle (1) and particle (2) should indeed always "be antiparallel"
to each other. A closer look at the quantum mechanical prediction (2.9c),
however, will soon lead us into some quite puzzling problems.
XIII.3 Bell's Inequalities, Hidden Variables, and the

Einstein-Podolsky-Rosen Paradox
According to (2.10), quantum mechanics predicts that the spin components

of the two particles (1) and (2) along a fixed direction a are always opposite
to each other. Instead of directly measuring ~(1) ® ](2) on particle (1) itself,
one may therefore determine the spin component of particle (1) along a
equally well by measuring ](1) ® IP) on particle (2) and multiplying the
result by - 1. This indirect "measurement" does not act on particle (1),
which in fact may be very far away from the applied apparatus; nevertheless,
it always yields the same value as a simultaneous direct measurement of
~(1) ® ](2) would. In view of this it appears quite natural to imagine that a
single particle (1) does not somehow" get" a definite a component of its
spin during a measurement of ¢(l) ® [(2), but rather already "has" a definite
value of it, either + 1 or - 1, prior to and independent of any measurement.
Moreovef, the direction a may be chosen arbitrarily except that it must be
orthogonal to the beam direction. Therefore one is led to the "natural"
assumption that prior to and independent of any measurement every single
particle (1) possesses definite values v(a), either + lor-I, for the com-
ponents of its spin, at least along all possible directions a orthogonal to the
beam. 11 These values are just "uncovered," rather than "produced," if
11 The cautious reader might notice already, however, that from the point of view of quantum
mechanics this hypothesis looks not so "natural" after all, since it ascribes simultaneously
fixed values to noncommuting observables.
actual spin measurements are performed. They may be visualized as hidden

labels attached to every single particle 0), one (either + 1 or -1) for every
possible direction a. The same argument applies, of course, to all particles (2).
Without an assumption of this kind it indeed appears quite difficult to
understand the perfect anticorrelation predicted by (2.10) for simultaneous
measurements of ~(1) @ ]<2) and /(1) @ ~(2). For if the value of ~(1) @ /(2) was
really undetermined until it is actually measured on particle (1), it would
appear impossible for particle (2)-which may be very far away-to" get
informed" about this value, in order to be able to "choose" just the opposite
value when /(1) @ ~(2) is measured on it. We will thus accept the hypothesis
formulated above, and use it to derive a simple inequality. It will turn out,
however, that this inequality is violated in practice!
Consider a very large number N of pairs of particles in the spin singlet
state (2.8) and four arbitrarily chosen directions a, b, c, and d in the plane
orthogonal to the two beams of particles produced by the source. Denote by
vla) and vld) the "hidden" predetermined values of the spin components
along a and d, respectively, of particle (1) in the ith pair, and by wi(b) and
wi(c) the "hidden" values of the spin components along band c, respectively,
of particle (2) in the same pair. Then, if the spin component of particle (1)
along a and the spin component of particle (2) along b are measured simul-
taneously for all N pairs, the results are vi(a) and wlb) for the ith pair, and
the average value (2.7c) becomes
1
E(a, b) = N I vi(a)Wi(b).
N
(3.1)
But one could as well have chosen other directions-e.g., d in the left and c
in the right beam-for the orientations of the two Stern-Gerlach devices. If
this experiment had been performed instead with the same N particle pairs, it
would have "uncovered" the spin components vi(d) and wi(c), instead of
vla) and wlb), and the observed correlation average (2.7c) would have been
1 N
E(d, c) = N i~l vld)wi(c). (3.1')
Analogous expressions may be written down for the results E(a, c) and
E(d, b) of two other conceivable experiments with the same N particle pairs. 12
Clearly at most one of these four alternative experiments can b~ actually
performed on the particle pairs i = 1, ... , N considered. Any actual experi-
mental determination of the four correlation averages E(a, b), E(a, c),
E(d, b), and E(d, c) thus requires (at least) four different sets of particle pairs.
However, which experiment is actually done on a given set of particle pairs
may be decided arbitrarily.
12 This is the" reproducibility assumption" of science: If an experiment is repeated it must

reproduce the same result. (We neglect statistical fluctuations, which is justified for sufficiently
large N.)
It seems reasonable to assume that an experimental determination of

E(d, c) would have produced the same result ifit had been performed on anyone
of the four sets of particle pairs. 12 In this sense E(a, b) ... E(d, c) are
quantities characteristic of the N particle pairs and their preparation, and
not dependent on which particular experiment is actually performed. In
particular, if quantum mechanics is true, both E(a, b) and E(d, c) as given
by the right-hand side of (3.1) and (3.1'), and also E(a, c) and E(d, b) when
expressed similarly, must agree with the corresponding quantum mechanical
predictions (2.9c), i.e., with - a . b, - d . c, - a . c, and - d . b, respectively.
At this moment, however, we will not yet assume this, but investigate
directly the" experimental" results given by equations (3.1), (3.1 ') and two
analogous formulas.
We want to derive an estimate for £(a, b) + £(a, c) + Bed, b) - Bed, c).
F or this purpose let us first show that
v;(a)(w;(b) + w;(c» + vi(d)(w;(b) - wi(c» = ±2. (3.2)
As Wi is + 1 or -1, the first bracket is either + 2, °

or - 2. If the first bracket
is + 2 or - 2, then the second bracket is 0, and if the first bracket is 0, the
second bracket is + 2 or - 2. As Vi is also + 1 or - 1, the left-hand side of
(3.2) is indeed either + 2 or - 2. Summing all expressions (3.2) over
i = 1 ... N therefore yields a number between - 2N and + 2N, and dividing
this by N we obtain
I~ (f v;(a)w;(b) + f vi(a)wi(c) + fVi(d)wi(b) - f Vi(d)W;(C») I s 2. (3.3)
By (3.1) and its analogs, the expression on the left is easily identified in
terms of £(a, b) ... Bed, c), and (3.3) becomes
IE(a, b) + E(a, c) + E(d, b) - E(d, c)1 s 2. (3.4)
This inequality is the most famous and experimentally most useful of a

series of similar estimates known as Bell's inequalities. 1 3
Let us now check whether the quantum mechanical prediction (2.9c) for
the spin zero state (2.8),
£(a, b) = (¢(l) ® #,2) = -a' b, (3.5)
satisfies Bell's inequality (3.4). Calculating the left-hand side of (3.4) using
(3.5) we obtain
1a • b + a •c + d •b - + c) + d • (b - c) 1
d • c 1= 1a • (b
s lallb + cl + Idllb - cl
= J2 + 2 cos cp + J2 - 2 cos cp,
131. S. Bell, Physics 1, 195 (1964). Various forms of Bell's inequalities are discussed in the
review by 1. F. Clauser and A. Shimony, Rep. Prog. Phys. 41, 1881 (1978). See also A. Peres,
Am. J. Phys. 46, 745 (1978).
a
c b c b
""------d d_----¥
a
Figure 3.1 Field configurations for which Bell's inequality is maximally violated.
<p being the angle between band c (b . c = cos <p). The last expression has a
maximum value 2J2 for <p = nl2 (cos <p = 0). Moreover, the inequality
sign in the above example becomes an equality for the following two
situations:
1. a and b + c, and d and b - c are parallel,
2. a and b + c, and d and b - care anti parallel.
These configurations of directions are depicted in Figure 3.1.
Therefore we obtain for the results E(a, b) of correlation measurements
as predicted by quantum mechanics the inequality
IE(a, b) + £(a, c) + £(d, b) - £(d, c)1 ~ 2vf2, (3.6)
and for the two configurations in Figure 3.1 this becomes an equality:
IE(a, b) + E(a, c) + E(d, b) - £(d, c)1 = 2~2. (3.7)
The latter, clearly, contradicts Bell's inequality (3.4), and there are indeed
infinitely many configurations of directions a, b, c, and d for which the
quantum mechanical predictions do not satisfy (3.4). (Any configuration
sufficiently" near" to the ones depicted in Figure 3.1 will do, the latter being
the configurations for which Bell's inequality is maximally violated.) Bell's
inequality (3.4) on the one hand, and the prediction (3.5) of quantum
mechanics on the other hand, are thus incompatible with each other, and
at least one of them must be wrong.
In this situation the best thing one can try is to decide the conflict
empirically. One simply has to do the experiment described in Section XIII.2
four times, with Stern-Gerlach devices oriented according to one of the
configurations in Figure 3.1, to determine the four averages £(a, b) ... £(d, c)
experimentally via (2.7c) (note that each of the four experiments consists of
many coincidence measurements on single particle pairs), to insert them
into the left-hand side of (3.7), and to look whether the result is ~ 2 in
accordance with (3.4), or equal to 2J2 (within experimental errors, of
course) as predicted by quantum mechanics. This is not as easy in practice
as one might imagine, however. Pairs of spin--! particles in the pure singlet
state (2.8) are difficult to prepare. In addition to that it has been shown that,
because of the uncertainty relations, Stern-Gerlach devices do not work
for charged particles,14 so that different (perhaps less precise) methods for
spin measurements have to be used for pairs of charged particles. Therefore
only one test of Bell's inequalities has been performed so far with spin-!
particles (protons). 15 The results agreed with quantum mechanics and
disagreed with one of Bell's inequalities, but the evaluation of the measure-
ments required additional hypotheses, and the results were not precise
enough to remove any doubts.
Much more precise experiments can be done with correlated photon pairs,
as produced by two-step (" cascade") transitions of atoms from a suitable
excited state to the ground state. In this case, the simultaneous spin measure-
ments considered above are to be replaced by simultaneous measurements
of the transverse linear polarizations of the two emitted photons along
arbitrarily chosen directions a and b. For the photon pairs emitted in
cascade transitions, these polarizations are correlated in much the same
way as the spin components of spin-! particle pairs in the singlet state (2.8).
A "natural" hypothesis analogous to the one made above again leads to
Bell's inequality (3.4) for suitably defined correlation averages E(a, b),
whereas quantum mechanics again yields the less stringent estimate (3.6).
The quantum mechanical prediction for E(a, b) is different from (3.5), how-
ever, and therefore the configurations of directions a, b, c, and d for which
(3.7) is true, and thus Bell's inequality (3.4) is maximally violated, are
different from those depicted in Figure 3.1. The results of recent experiments
of this kind 16 are in excellent agreement with quantum mechanics, thus
proving beyond any doubt that Bell's inequality (3.4) is not respected by
nature. Forced by this unequivocal verdict against (3.4), we have to renounce
the apparently "natural" hypothesis made above from which this wrong
result (3.4) was derived. We must realize that it was inconsistent to assume
that a particle can simultaneously "have" fixed values of its spin com-
ponents along different directions, regardless of whether or not they are
really measured. The value of a spin component that has not been measured
is not just unknown-it does not even exist. Imaginary experiments do not
have results at all, not even imagined ones. 1 7
Looking back at Equation (3.2), which was the crucial step in proving
(3.4), we see that it indeed contains values of spin components v;(a) and
v;(d) (or w;(b) and w;(c)) which have not been measured simultaneously on
the same (the ith) particle pair, but were only imagined to be simultaneously
present on the basis of our" natural" hypothesis. The only way to avoid the
14 N. F. Mott and H. S. W. Massey. The Theory of' Atomic Collisions, 3rd ed., Section IX.2.
Clarendon Press, Oxford, 1965.
15 M. Lamehi-Rachti and W. Mittig, Phys. Rev. 014,2543 (1976).
16 A. Aspect, P. Grangier, and G. Roger, Phys. Rev. Letters 49, 91 (1982); A. Aspect,
1. Dalibard, and G. Roger, Phys. Rev. Letters 49, 1804 (1982).
17 "No elementary quantum phenomenon is a phenomenon until it is an observed
phenomenon." (1. A. Wheeler.)
wrong but inevitable consequence (3.4) of (3.2) is indeed to admit that some
of the quantities in (3.2) must be meaningless. If v;(a) and wi(b) are really
measured values, Equation (3.1) is the correct formula for the correlation
average measured in a real experiment. But then vied) and wi(c) simply do
not exist, and Equation (3.1/) is completely meaningless. A real experimental
value of E(d, c) used in the test of (3.4) involves particle pairs from a different
set of pairs, and (3.1/) should be replaced by
1 M
E(d, c) =- L: va(d)wa(c) (3.1//)
M a=l
with the measured values vaCd) and wa(c), where the index !X. enumerates this
different set of pairs. But there is no way to derive Bell's inequality (3.4)
from (3.1), (3.1//) and two similar equations for E(a, c) and E(d, b).
The hypothesis that the two spin-t particles "have" definite spin com-
ponents v;(a) and w;(b) seems natural only from the classical point of view,
in which one visualizes the particle pair to consist of two separate particles.
However, the quantum mechanical state vector cjJ of (2.8) does not describe a
state with separate single-particle properties. Such states would be described
by the basis vectors (2.3) or (2.3/), which however are not eigenvectors of
(S(l) + S(2»2, i.e., have no well-defined total spin. The state cjJ describes a
new entity, an indivisible whole, a single object whose constituent particles
(1) and (2) are not definable until a measurement is made that prepares the
direct product states (2.3/) (or mixtures thereof). As a state with total spin
zero, this state cjJ does not have single-particle properties. And as a pure
state it cannot be subdivided further, not even in our minds.
In quantum mechanics, due to the superposition principle, one can obtain
new states for the combination of two subsystems (basic postulate IIa)
which no longer have single-constituent properties like the direct product
states (2.3/). These direct product states exist too, but in addition there is
the possibility of a multitude of linear combinations, of which cjJ is one
example.
In classical physics the building blocks of a composite system are its
constituents. In quantum physics the building blocks are the subspaces of the
space of physical states (" Hilbert space"). They can be any kinds of subs paces,
not only the one-dimensional subspaces spanned by the direct product
vectors of constituent states. The state cjJ is one of this multitude of possible
states.
Historical Remark
The invalidity of the consequences of apparently" natural" ideas, like the
inequality (3.4), is one of the numerous "counterintuitive" features of
quantum mechanics. Such features made it difficult for many physicists, who
grew up in the classical tradition, to accept the interpretation of quantum
mechanics that evolved in the late 1920s and early '30s. Particularly famous
are the arguments of the 1935 paper by Einstein, Podolsky, and Rosen (EPR)
who describe essentially the same "paradox" as the one discussed above.
They considered a pair of point particles with coordinates (one-dimensional,
for simplicity) Xl and X 2 in the (improper) state ¢ described by the wave
function
(3.8)
The momentum-space representation of this state ¢ is the Fourier transform

of (3.8), cf. (II. 7.49),
<PIP21 ¢) = 2~ J
e-i(P1Xl +P2 X 2)<X I X 2 1¢) dXI dX2
(3.9)
According to (3.8) simultaneous measurements of the positions Ql and Q2

of the two particles always yield two values Xl and X 2 related by XI = X 2 + a,
and according to (3.9) simultaneous measurements of the momenta PI
and P 2 always yield two values PI and P2 with PI = - P2' Therefore either
the position Xl or the momentum PI of particle (1) can be determined by a
corresponding position or momentum measurement on particle (2).
Although these two measurements on particle (2) cannot be performed
simultaneously, neither of them acts directly on particle (1). From this
EPR conclude-as we did above in the analogous case of spin measurements
-that such measurements cannot "produce" the measured values XI and
PI but merely "uncover" them. Therefore both the position and the
momentum of particle (1) should have some kind of "physical reality,"
independent of whether or not they are actually measured. However, accord-
ing to the uncertainty principle, quantum mechanics cannot simultaneously
ascribe to a particle definite values of both position and momentum!
EPR conclude from this example that quantum mechanics does not provide
a complete description of physical reality.
Bohr, in a reply published later in 1935, tried to defend quantum
mechanics against this criticism, but apparently did not convince everyone.
Since then and until quite recently some physicists have looked for theories
which are "more complete" than quantum mechanics in the following
sense: One assumes that there exists a more precise specification of the states
of a micro system in terms of additional "hidden" variables, such that in
the new" microstates" -as specified by fixed values of the hidden variables
-all observables simultaneously possess fixed values. The uncertainty
relations of quantum mechanics are therefore not valid in these" microstates."
On the other hand, the uncertainty relations of quantum mechanics are
fulfilled even in a pure quantum state (cf. Chapter II). A pure state of quantum
mechanics is therefore interpreted in the new theory as a mixture of the
more precisely defined microstates of the system. Any uncertainties in the
outcomes of measurements on a single system in a given pure quantum
state then merely result from our ignorance of the" real" microstate of the
system considered, which could be any of the microstates which are mixed
up in the pure quantum state. This is the uncertainty of classical physics,
which comes from not knowing enough and can in principle be overcome;
it is familiar from classical statistical mechanics. Thus the novel quantum
mechanical uncertainty of "not being able to know" would be converted
by a hidden-variables theory into the more familiar uncertainty of "not
yet knowing."
While preparing a review article on hidden-variables theories, Bell
discovered in 1964 that such theories almost inevitably lead to certain
estimates-now known as Bell's inequalities-which are not satisfied by the
predictions of quantum mechanics. Indeed the above derivation of the
particular example (3.4) of Bell's inequalities used almost nothing but the
hypothesis that "in reality" certain spin components of a single spin-!
particle-whether or not they are actually measured-have fixed values. This
hypothesis follows directly from the assumption that hidden variables
in the sense explained above exist, independently of any additional details
of the theory. 18
Bell's inequalities were a decisive step towards resolving th~ controversy.
They moved the question of hidden variables from the realm of pure theory
into the realm of experimentally decidable propositions. They also provided
an opportunity to falsify quantum mechanics. If the experimental data
would have fulfilled Bell's inequality rather than the predictions of quantum
mechanics, then quantum mechanics would have turned out to be not only
incomplete (as suggested by Einstein and collaborators) but even wrong. By
confirming the quantum mechanical predictions, however, the crucial
experiments have not only supported quantum mechanics but also ruled
out once and for all the whole class of local hidden-variables theories.
Problems
1. Let a be the vector which is obtained from the vector e 3 (z-direction) by a rotation 9.
Let Ia ±) be the eigenvectors of a· S with eigenvalues ±!, and I ±) the eigenvectors
of S3 = e 3 • S. Prove the relation
Hint: Use that in ~F 1;2 the operators Si fulfill the anticommutation relation {Si, SJ =
!biJ.
2. Using the equations (3.23), (3.22), and (3.8) of Chapter III or the Pauli matrices
of Problem 5, Chapter III, show that the matrix elements of the operator Ii = 2a . S
18 We did assume in addition, however, that the "hidden" spin components of particle (1)
are not influenced by spin measurements on particle (2), and vice versa. Hidden-variables
theories with this property are called local. Bell's inequalities can be derived for local hidden-
variables theories only.
Problems 355
between the eigenvectors Is = ~ S3 = ±} >= I ± >of S 3 are given by

<± I~I ±> = ±a3,
<± l(hl =+=> = a =+= ia2,
1
where ai = a . ei are the components of the vector a.

3. Using the results of Problem 2 show that the expectation values of the operators
¢(1) ® [(2\ [(1) ® ~(2) and ¢(l) ® ~(2), defined in equations (2.4)-(2.6), in the state
¢ of (2.8) are given by <4>I¢(1) ® [(2)14» = 0; <4>11(1) ® W2)14» = 0; E(a, b) ==
<4>I¢(1) ® ~(2)14» = -a' b.
CHAPTER XIV
Transitions in Quantum Physical

Systems-Cross Section
This is the fundamental chapter on scattering theory. Here the concept of

transition probabilities is introduced and general formulas for the cross
section are derived from the basic assumptions of quantum mechanics. In
Section XIV.2, a general formula for the transition rate is obtained, which
will be used in Section XIY.S for the derivation of the cross-section formula
and in Chapter XXI for the derivation of the decay rate. Section XIV.3 intro-
duces the concept of cross section, and Section XIV.4 gives a brief description
of the different ways to relate the cross section to fundamental physical ob-
servables. The main section of this chapter is Section XIV.S, where the cross-
section formulas for very general physical situations are derived. At the end
of section XIV.S the physical state of the scattering system is further specified
to obtain the well-known expressions for the cross section. For a superficial
understanding of the material of this chapter one may omit parts of Section
XIV.2 and Section XIV.S. For this purpose we have discussed special results
of Section XIV.S at the end of Section XIV.4.
XIV.1 Introduction
In the present chapter we collect some material that will be used in subse-
quent chapters to describe collision and decay processes. We derive the
transition probability, introduce the notion of transition rate, and then define
the cross section and derive the formula that relates it to the transition
probability.
356
XIV.! Introduction 357
The general situation that we shall describe, and of which collision and
decay processes are just two particular cases, is the change in time of a
nonstationary state Wet). We shall use the basic assumptions of quantum
mechanics, in particular the axiom V. We shall make the further simplifying
assumption that the generator H of time translation can be split into two
parts:
H=K+V, (Ll)
where K is the energy operator of the isolated physical system with stationary
states and is assumed to be well defined. This means that it is assumed to
make sense to consider an approximate description of the physical system
by a stationary physical system. There are many examples for which this is
possible. Thus K may be the energy operator of an atom or of a molecule.
In the approximate description the states belonging to an energy level are
stationary, but in reality they are only quasistationary and decay into the
ground state. This transition is then caused by V = H - K. Another example
is the combination of two physical systems that can be spatially separated far
apart from each other, a situation which one usually encounters in collision
processes. If they are far apart, the energy operator of the combined system
is K; and V = H - K, which describes the interaction between the two
subsystems, has a finite range. 1
The observables that are measured in these processes are assumed to
commute with the operator K. For example, a detector may register all the
ground states of atoms that were initially in excited states (by registering,
for example, all emitted light of a particular frequency). The observable
measured is thus the projection operator on that subspace of the space of
physical states which contains the ground states of the atoms, i.e., the ob-
servable measured is the property of being in the atomic ground state (cf.
Section IIA). In the other example of collision experiments the detector is
placed far away from the target and therefore detects eigenstates or mixtures
of eigenstates of the operator K. The property B measured by the detector is
described by a projection operator A B , which projects onto a (in general
continuous) direct sum of energy eigenspaces of K. Usually B is a more
specific property. For example, the detectors may not be placed all around
the target, but only at a particular angle. AB is then the projection operator
onto that particular subspace of the above direct sum of energy subspaces that
contains states whose momentum vectors are directed into the particular
solid angle.
Thus the problem that we shall discuss (in the Schrodinger picture) is the
following: The state Wet) changes in time according to
(W == W(O)). (1.2)
1 It should be remarked that the assumption (l.l) is not really necessary for the description
of collision processes; it suffices to assume only that asymptotic direct-product states do exist.
358 XIV Transitions in Quantum Physical Systems-Cross Section
What is the expectation value of an observable A (i.e., what is the probability

of measuring the property A) for which
[A, K] = O? (1.3)
This problem is often interpreted as the transition between different sta-

tionary states of a physical system caused by an interaction V. This is based
on the assumed existence of a K eigenstate win(t = - CfJ) in the remote past
and of a K eigenstate wout(t = + (0) in the far future that equals or con-
tains A.
XIV.2 Transition Probabilities and Transition Rates

Denote the state of the physical system by W, and the observable to be
measured by A. Then, according to the basic postulate II, the expectation
value <A) of A is given by
<A) = Tr(AW). (2.1)
Since A is a projector, <A) is the probability that the system is in the sub-
space AYf. <A) is called the transition probability into the state A.
The transition probability is a function of time. At the beginning of the
process <A) will in general be zero, and will then increase with time. Ac-
cording to the results of Chapter XII, this time dependence can be described
in various pictures. We shall use the Schrodinger picture first, because it
appeals best to our intuitive understanding of a scattering experiment with
a pulsed beam or of a decay process. In the Schrodinger picture the ensemble
of physical systems is described by the time-dependent statistical operator
Wet), which describes the beam moving towards the detector, or the excited
state of an atom decaying into the ground state.
The observable A, which represents the apparatus, does not change in
time. [We assume that A does not depend explicitly upon time (oA/ot = 0)
i.e., that the apparatus is not influenced from the outside after the scattering
experiment has begun.] Thus the transition probability is given by
<A\ = Tr(AW(t», (2.2)

where
Wet) = ut(t)WU(t) (W == W(O»;
W is the state of the system at some conveniently chosen time, which we
call t = 0, and U(t) = eiHt/li is the time-evolution operator of Chapter XII.
The transition rate is defined to be the time rate of change of the transition
probability: d<A)t/dt. Differentiation of (2.2) and use of (XII. 1.33) give
d <A)t = Tr ( A ~
dt dW(t») = - i Tr(A[H, Wet)]). (2.3)
Substitution of H = K + V into (2.3) yields

d
dt <A)t = -i Tr(A[V, Wet)]) - i Tr(A[K, Wet)]). (2.4)
With the use of (1.3) the second term in (2.4) is easily shown to vanish:
Tr(A[K, Wet)]) = Tr(A[K, Wet)] + [K, A]W(t))
= Tr(AKW(t) - AW(t)K + KAW(t) - AKW(t))
= - Tr(A W(t)K) + Tr(KA Wet)) = O.
Let {I b)} be a basis for the subspace AYe onto which A projects. Then
Tr(AW(t)V) = L <bl W(t)vlb) = L <bIVW(t)lb)*
b b
= (Tr(AVW(t)))*.
The right-hand side of (2.4) may then be written
d
dt <A)t = - i Tr(AVW(t)) + i(Tr(AVW(t)))*. (2.5)
But i(lX* - IX) = 21m IX; consequentlyz

d
h dt <A\ = 21m Tr(AVW(t)). (2.6)
In order to express the transition rate in a form more easily used, we use a
basis of generalized eigenvectors of K and of H:
Kia) = Eala), (2.7a)
and
(2.7b)
Recall that the generalized eigenvectors Ia ±) of H are related to the general-
ized eigenvectors Ia) of K by means of the Lippmann-Schwinger equation
[Equation (VIII.2.7')]:
(2.8)
On occasion we shall make the labeling on Ia) and Ia ±) more explicit by

writing
(a) la) = IEaa) and (b) la±) = IEaa±),
where a = (aI, az, ... ,ak) are labels needed in addition to the energy to
specify the generalized vectors. These labels (quantum numbers) are eigen-
values of additional operators A 1> A z, ... , Ab which together with K form
a c.s.c.o. For example, a may consist of the angular-momentum quantum
2 We use the units with h = I and introduce only in a few important formulas the correct
factors of n.
numbers 1and 13 together with other, internal quantum numbers '1; or amay
consist of the direction p/l p 1 of the momentum ptogether with other quantum
numbers '1. If the operators Ai (i = 1,2, ... , k) all commute with H, then the
1 Eaa±) are generalized eigenvectors of the C.S.C.o. {H, A 1, ... , A k }; other-
wise the additional labels a just serve to indicate that 1 Eaa±) is obtained
from 1Eaa) by way of the Lippmann-Schwinger equation, and do not indicate
that 1Ea a±) is an eigenvector of all the A;'s (for more details cf. Section XV.1).
We shall assume that the K eigenvectors 1a) are normalized according to
(2.9a)
[p(Ea) is a weight function that is arbitrary but fixed 3 ; a convenient choice

for p(Ea) will be made later.] With such a normalization the summation
La is really an abbreviation for the more explicit
La = JP(Ea) dEa 4:.

a
(2.10)
The generalized eigenvectors 1 a ±) of H have the same normalization

(2.9b)
as the corresponding generalized eigenvectors a) of K. The derivation of

1
(2.9b) from (2.9a) is given in Appendix XV.A.

In general the spectrum of K and the spectrum of H are not the same.
Usually K has only a continuous spectrum {Ea} starting at a particular
value, say E = 0, and going to infinity. But the spectrum {Eo} of H is the
combination of a continuous spectrum {Ea}, which is usually the same as
the continuous spectrum of K, and a discrete spectrum {En}, which may be
negative but is bounded from below. Physically the continuous spectrum
corresponds to the scattering states and the discrete spectrum to the bound
states of the projectile and target. There may be discrete eigenvalues in the
continuous spectrum, but for physical reasons the energy eigenvalues can-
not be arbitrarily large-in particular, not arbitrarily large and negative. We
may thus choose the generalized eigenvectors of K:
as a basis for the space of physical states, or we may choose the set
of discrete eigenvectors 1an) and generalized eigenvectors 1a ±) of H as a

basis. The completeness property of these bases may be expressed as
1= L la)<al = JP(Ea) dEa L IEaa)<Eaal (2.l1a)

a a
3 It is for the generalized energy eigenvectors IE) what p(x) - 1 is for the generalized eigen-
vectors Ix}p in (1.4.14) and (1.4.7c p ).
and as
I = ~ la)<al = f p(Ea) dEa ~ IEaa±)<Eaa± I + ~ lan)<anl· (2. 11 b)
The trace that appears on the right-hand side of (2.6) may now be written
Tr(AVW(t» = I <bl VW(t)lb)
b
= I I <blVla)<al W(t)lb).
b a
Use ofthe Schr6dinger-picture time development (1.2) of W(t) and insertion

of another complete set of H eigenvectors la') then gives
Tr(AVW(t» =I I (bIVla)(ale-iHtWe+iHtla')(a'lb)
b Clrz'
= I I (bl Vla)(al Wla')(a'lb)e-i(E~-E~')t. (2.12)

b CXCl'
Suppose H does have discrete eigenvalues, corresponding to bound states

Ian). Since bound states cannot evolve from free states, we assume that the
state W contains no contributions from bound states; mathematically this
assumption is expressed by
(al Wla~) = O. (2.13)
We may therefore restrict the summations over a and a' to the continuous
spectrum of H.
Since [A, K] = 0, the basis {Ib)} of the subspace Aye may be chosen to
consist of generalized eigenvectors of K:
(2.14)
If we take the inner product of the Lippmann-Schwinger equation (2.8)

with Ib), we obtain with (2.14)
(a' ± Ib) = (a' Ib) + lim (a' ± I V 1 . Ib)

<-+O± Ea , - K + IE
Use of (2.15+)4 in (2.12) yields
Tr(AVW(t» = I Ie -i(Ea-Ea')1 (blVla+) (a'+ I.Vlb) (a+ IWla'+) + II,

baa' Ea, - Eb - 10
(2.16)
4 That it is more natural to choose £ _ 0 + rather than £ _ 0 - can be seen from Section
XV.3, where we discuss in detail the significance of the choice £ - 0+. However the derivation
could have as well been continued with (2.15_) and would lead to the same results.
362 XIV Transitions in Quantum Physical Systems~Cross Section
where
II = I I e-i(Ea-Eu,)t<bl Vla+)<a'lb)<a+ IWla'+), (2.17)
baa'
In Appendix XV, A we shall show that the second term II vanishes; we con-
tent ourselves here with an intuitive argument: The factor <a' Ib) in the
second term does not describe a transition, but describes the probability of
observing the configuration of the initial state, But in a scattering experiment
(except for absorption measurements, which do not measure <A) directly),
the detectors are placed at an angle to the direction of the incident beam so
that they are not flooded by the incident beam. Thus if lEA 1) = IA)
denotes any of the vectors that appear in the initial state W = I W A IA)<A I,
then I A) f: AYf and the term II does not contribute to the probability of
observing A. 5
The transition rate is then obtained by inserting the first term of (2,16)
into (2,5). By using the Hermiticity properties
<a+ I Wla'+)* = <a'+ I Wla+) and <bl Vla+)* = <a+ I Vlb)
of Wand V, and by exchanging the summation indices a and a' in the second
term arising from (2.5), one obtains
x (E a, ~ ~b ~ iO - Ea - ~b + iO). (2.18)
We will make use of this last result both in Section XIV.5 when we calculate
the cross section and in Section XXI.2 when we calculate the decay rate.
XIV.3 Cross Sections
In a scattering experiment a beam is directed towards a small target. A

detector is located at a large distance from the target and at a (nonzero) angle
n = (8, c/J) with respect to the incident beam.
In classical physics this beam can be a beam of particles or a beam of
radiation. For the case of particle scattering, for example, a beam of N B
particles of type B with mass mB and velocity Vo is, during an interval of time
L'lt, senttowards a target consisting of N T particles of mass mT' The number of
particles per second that are scattered and reach the detector, N/L'lt~the
counting rate, is proportional to both the number NT of particles in the
target and to the incoming flux (i.e., to the number of particles per second
per unit area perpendicular to vo). The incoming flux is given by
incident flux = PB VO , (3.1)
5 Note that" is the term that comes from the first term in the Lippmann-Schwinger equation.
XIV.3 Cross Sections 363
Figure 3.1 Schematic diagram of a scattering experiment.
where PB is the density of particles in the beam. Thus the counting rate
NjAt is
(3.2)
where the proportionality constant a is called the cross section. The cross
section for the scattering of particles by particles is thus defined by
NjN T
a - (3.3)
- PBvoAt"
It has the dimensions of an area. [If the targ~t is not at rest, then the relative
speed v = IVo - VT I should be used in pla~e of Vo in (3.2).]
Cross sections are usually measured in barns (1 b = 10- 24 cm 2 ), millibarns
(1 mb = 10- 27 cm 2 ) or micro barns (1 pb = 10- 30 cm 2 ). One can obtain
an intuitive understanding of the definition (3.3) for the cross section a in
the following way: The number of particles per unit time N j At which
reach the detector is eqwil to the number of" particles per unit time
PB(AzjAt)a, which pass through the volume Aza times the number of target
particles NT:
a is thus the surface area perpendicular to the incident beam which the target
particle presents to the beam particle. Over how large a surface the target
particle affects the beam particle depends of course upon the nature of these
two particles, in particular upon their interaction. But the order of magnitude
of a is roughly defined by the" size" of the interacting particle or the range of
the interaction. Typical cross sections for strongly interacting particles
are therefore of the order of millibarns and microbarns, and cross sections
in atomic physics (elastic scattering) are of the order of 10- 16 cm 2 .
Different kinds of cross section are given different names. If the detector
detects all particles of a third kind that leave the target area in all directions,
then a is called the production cross section. If the detector detects particles
of type B, the incident type, after they leave the target area, then a is called
the scattering cross section. In particular a is called the total scattering cross
section if the detector detects all particles B leaving the target area without
regard to the direction in which the particles leave. The differential cross
section da is obtained if only those dN particles which are directed into a
specific cone of (infinitesimal) solid angle dO. = sin e de d<jJ are detected
cf. Figure 3.1):
da = dN jAt da
or dO. (3.4)
PBNTVO
The other example of a scattering experiment in classical physics is the

scattering of a beam of radiation incident on an obstacle (target). The
detector, which is located outside the incident beam at a distance large
compared with the wavelength of the radiation and with the size of the ob-
stacle, measures the flux scattered in a given direction. The ratio of the
differential scattered power per unit area dI scat to the incident power per
unit area (flux) Iincid is the differential cross section
d(J = dI scat . (3.5)

I incid
In quantum physics the beam is a beam of quantum physical systems, and

the measured quantities are probabilities. The detector measures the proba-
bility of a transition from the incident state. This transition probability is
proportional to the probability that a particle before scattering is passing
through a surface of unit area perpendicular to the incident velocity. The
proportionality constant is the quantum-mechanical cross section:
transition probability
(J= ~~----~~~------~---- (3.6)
incident probability per unit area·
As in the classical case, there are different kinds of cross section. If the
detector registers all states that emerge from the target, then (J is the total
cross section. Ifthe detector registers only a subset of states (e.g., only those
states which are in a state B), then (J = (JBA is called the partial cross section.
Thus
transition probability from a state A into a state B
(JBA = (3.7)
incident probability per unit area
Suppose B is the property that the system's momentum vector is directed

into a specific cone of solid angle dQ around the direction Q = (8, 1jJ); then
(transition probability from a state A into

d _ all states with momentum vector directed into dQ)
(3.8)
(JnA - incident probability per unit area '
or more often d(JnA/dQ is called the differential cross section.

In typical scattering experiments the projectile is chosen to be as structure-
less as possible (e.g., electrons), whereas the target consists of more com-
plicated objects (e.g., atoms). The experimental cross sections then reveal
information about the structure of the target particles. The collision of the
projectile with the target can be elastic, as is the case when the energy of the
projectile is smaller than the difference between the internal energy levels
of the target; or the collision can be inelastic, as is the case when the internal
energy level of the target is changed by the collision. To fix the nomen-
clature we shall use "total collision cross section" for the cross section (J
of a collision that involves any elastic or inelastic process. The cross section
XIVA The Relation of Cross Sections to the Fundamental Physical Observables 365
for all elastic collisions is called the total elastic cross section while the
(J elas
cross section (J~ for inelastic collisions into the 1]th internal energy level will be
called the total inelastic partial cross section for the 1]th level. The total
inelastic cross section is L~ (J ~. The total cross section (J is then the sum of
the total elastic cross section with the total inelastic cross sections:
(3.9)
Corresponding notation is used for the differential cross section.
XIV.4 The Relation of Cross Sections to the Fundamental

Physical Observables
Cross sections are the observable quantities. Their measurement follows in

principle from their definition, given in the previous section, in terms of
counting rates and flux. 6 The cross sections are related to the fundamental
observables which describe the structure of the physical system consisting
of the combination of projectile and target. We shall now establish the con-
nection of the cross sections to these quantum-mechanical observables.
If the theory of the physical system (i.e., the mathematical structure that
describes the system) is known, then one can calculate the cross sections in
terms of the known fundamental observables and predict the outcome of a
collision experiment. More often a physicist meets the reverse situation:
The cross sections have been measured experimentally; from this informa-
tion conjectures are made as to what the fundamental observables are, how
they behave, and what mathematical structure best describes such behavior.
This task is least complicated if one of the colliding subsystems (called the
projectile) has a known structure that is as simple as possible. Such a pro-
jectile serves to test the structure of the target system and to provide informa-
tion for a theoretical model of the target system. Some examples are electrons
as projectiles with atoms or molecules as targets, protons as projectiles with
nucleons as targets, and electrons or other leptons as projectiles with hadrons
as targets.
To relate the cross sections to fundamental observables one may proceed
in two different ways. The first and conventional way is based upon the time-
development axiom V of Chapter XII. It is assumed that there exists an energy
operator H in the system's algebra of operators that develops the system in a
continuous fashion. It is usually further assumed that the energy operator
H can be split into parts,
H=K+V, (4.1)
6 The actual measurement of cross sections is described in books on experimental physics.

See. for example. Massey et al. (1969). Vol. 1.
where K is the energy operator of the combination of the physical systems

in the absence of the interaction V between them, KB is the energy operator
for the beam system, and KT is the energy operator for the target system.
K is thus the energy operator when the projectile and target are far apart. 7
The existence of a continuous unitary time development (1.2), generated
by an operator H according to the basic assumption V, is a rather question-
able assumption in particle physics, where a fundamental length is believed
to exist. Therefore a second way was suggested by Heisenberg (1943). It is
based on the realization that Wet), which describes the systems while they
are interacting, is really not a measurable quantity. Measurable quantities
are the initial state Win(t ---> - CIJ) that describes the systems before the
interaction when they are prepared for the collision process, and the final
state wout(t ---> + CIJ) that describes the systems after the interaction when
they are detected. Each initial state is transformed into a final state by the
interaction. The operator that describes this transformation
or, for pure states,
is called the S-operator (" S-matrix ").

S is to have the following properties: (1) It transforms superpositIOns
a<l>in + b,¥in into superpositions a<l>°ut + b'¥out; therefore it must be a linear
operator. (2) It transforms every normalized initial state uniquely into a
normalized final state (" conservation of probability"); this, together with
the assumption that the set of initial states as well as the set of final states
spans the space of physical states, requires that S be unitary: sst = sts = 1.
The cross section can then be related to the matrix elements of the op-
erator S (S-matrix). Thus in this second approach the existence of a unitary
S-operator is taken as the fundamental postulate in place of the basic as-
sumption V of a continuous unitary time development.
If the Hamiltonian H exists and is the generator of a continuous time
development, then the S-matrix may be expressed in terms of H,8 The S-
operator is still a meaningful concept, however, even if this is not the case.
Heisenberg's assumption that it is the S-operator (and not the Hamiltonian)
that is the fundamental physical quantity has become the basis for an attempt
at a new formulation of elementary-particle theory known as S-matrix
theory.
The first way thus assumes more and allows one to relate the observable
quantities, such as cross sections, to fundamental observables that reveal
more of the physical system's structure. The second way denies the pos-
7 The assumption (4.1) that H can be split is not really essential. One can derive expressions
for the cross sections from the assumption of the existence of a time-development generator H
alone. See. for example, Goldberger and Watson (1964), Chapter 5.
8 In this context the S-matrix was first introduced by J. A. Wheeler in 1937.
XIV.4 The Relation of Cross Sections to the Fundamental Physical Observables 367
sibility of this deeper insight into the structure of the physical system. For
atomic scattering experiments in nonrelativistic quantum mechanics, the
existence of an elementary length of less than 10- 13 cm (= 10- 5 A) can be
ignored. We can therefore connect the cross section to our basic assumptions
for quantum mechanics (in particular to the axiom V) by proceeding in the
first way, along which we shall meet the S operator as a derived quantity.
In the next chapter we shall examine in more detail the concept of in-
coming and outgoing states and their relationship to the S-matrix.
In Section XIV.5 we shall give a derivation of the cross section from the
basic assumptions of quantum mechanics under very general conditions.
Here we list a few special results for the benefit of the reader who does not
want to go through the tedious derivations of the following section.
The cross section is expressed in terms of the matrix elements of the
interaction Hamiltonian V or the T-matrix. The matrix elements of V and
T are related by
(5.41')
Here E = p2/2m + E~ denotes the energy of the target-projectile system,

E~ denotes the internal energy (which may be a function of the internal
quantum numbers 1]), and 0 = (8¢) denotes the direction of the projectile
momentum, P = po.. lEO I] +) is the eigenvector of the exact operator H,
>
and lEO I] is the eigenvector of the free-energy operator K = H - V. If a
Hamiltonian time development does not exist, then the T-matrix is con-
nected with the S-matrix by
<EOI]ISIE'O'I]') = <EOl'/I£,O'I]') - 2nib(E - E')<EnYfITIEn'Yf').
(XV.3.36')
If the generalized eigenvectors are normalized according to
then the differential cross section for the scattering of a projectile with mass
mA and momentum PA on a target in a state I]A into a particle with mass mb
going into the direction fib = Pb/Pb and leaving the target in the new state
I]b, is given by
d(J (2n)4h 2 mAmbPb(E A) 2

dn(EAnbl]b~PAI]A)= PA1<EAnbl]bITlpAI]A)I, (5.61')
where
p~
E
A
= ----
2mA
+ E ~A'
For elastic scattering, mA = mb, I]A = I]b, this goes into (5.62) at the end of
Section XIV.5, which can be expressed in terms of the elastic-scattering
amplitude (5.63) given by (5.64).
XIV.5 9 Derivation of Cross-Section Formulas for the

Scattering of a Beam off a Fixed Target
We shall now give the precise expression for the quantum-mechanical de-
finition of the cross section in terms of the state and the transition probability.
We shall then proceed to derive various cross-section formulas.
The transition probability per unit time, corresponding to d(N/NTNB)/dt
for classical particle scattering, is in quantum mechanics given by d(A\/dt.
The probability that the system will make a transition during the time period
tlt taken by the experiment is (d(A)I/dt) tlt.
The probability density for the beam position, corresponding to PB/ N B
for classical particle scattering, is given in quantum mechanics by
(x I W~(t) Ix), where W~(t) is the statistical operator describing the state
of the incident beam system. If we assume a small spread in the velocities
of the beam's constituent particles, the incident probability per unit area
per unit time for a beam that is incident in the z-direction with an average
velocity Vo is (x I W~(t) Ix)vo'
According to its definition (3.7) and corresponding to its classical ex-
pression (3.3), the cross section for an experiment that runs over the period
tlt is then given as the proportionality factor a in the equation
d~~)1 tlt = a(x I W~(t) I x)vo tlt. (5.1)
If the experiment extends over a long time period, the defining equation (5.1)
goes over into the equation
f + dt d(A)
00
_ _I =
f+ 00
dt a(x I W~(t) Ix)vo' (5.2)
- 00 dt - 00
If the beam does not consist of structureless quantum physical systems

but has internal degrees of freedom described by a set of quantum numbers
A, then the beam's position probability density is LA
(x AI W~(t) Ix A), where
Ix A) is a basis of generalized position eigenvectors for the space of physical
states of the beam system. The quantum numbers A usually describe the
polarization (spin state), and we shall therefore generally refer to A as the
polarization, even though in the case of more complicated projectiles (e.g.,
atoms, molecules, or ions), Acould stand for a whole set of quantum numbers
upon which the energy eigenvalues of the beam system may depend.
If there are polarizations, then one has
+OO d(A) f+oo .
f -00 dtT= -00 dtavo~(xAIW~(t)lxA) (5.3)
instead of (5.2).
9 To follow these derivations may require special effort. We have therefore presented a few
special results needed for the following chapters at the end of Section XIVA and this section
may be omitted in first reading.
XIV.5 Derivation of Formulas for the Scattering of a Beam off a Fixed Target 369
We assume in this section that the target is fixed in position; Vo is thus the
average relative speed between projectile (beam) and target. We further
assume that in the absence of the interaction V between the component
systems, the statistical operator win(t) for the combined projectile-target
system would be factorable into the product
(5.4)
of a statistical operator W~(t) describing the state of the projectile with a
statistical operator W!;(t) describing the state of the target. W~(t) acts in a
space of states YfB, W~(t) acts in a space YfT , and Win(t) acts in the total
space Yf = YfB ® YfT ·
The justification for these assumptions lies in the initial conditions:
Win(t) can be thought of as the operator which describes the system before
the interaction becomes effective (i.e., for t --+ - 00), for it is then that the
beam and target system are prepared. Developing in time according to
where K is the interaction free energy operator, it describes a fictitious state

in which the beam passes through the target without interaction. The pre-
paration of the component systems is by way of measurements of properties
which are uncorrelated at the time of preparation, with the consequence
that at the time of preparation Wln(t) is factorable into the form (5.4).10 But
then it remains so at all times, as follows from:
Win(t) = e-iiKB+KT)t(W~ ® W!;)e+i(KB+KT)t
= e-iKBtW~eiKBt ® e-iKTtW!;eiKTt
= W~(t) ® W!;(t). (5.5a)
The actual state of the combined system is Wet), which, while it is es-
sentially the same as Win(t) in the distant past, has a different time develop-
ment [Equation (1.2)J:
(5.5b)
Unlike win(t), Wet) is generally not factorable, because the interaction part
V of its time-development generator H = K + V acts non trivially in both
YfB and YfT .
Rather than interrupt the calculations to follow with many explanations,
we shall first make some remarks about the basis vectors which will be used
and their normalizations, and shall then list some needed mathematical
relations.
For the basis {I E 8)} of Yf = YfB ® YfT we shall use the direct-product
basis consisting of the generalized K eigenvectors
IE 8) = IpA) ® IET~)' (5.6)
Here Ip A) is a generalized eigenvector with polarization A of the momentum
10 Cf. remarks in Section XIII.1 concerning the state of a combination of physical systems.
operator P for the beam system, and IETI]) is a (possibly generalized) eigen-
vector with polarization I] of the operator KT = K~t for the internal energy
of the target systemY The generalized eigenvectors Ip A) have the usual
b-function normalization,
(S.7)
while the IETI])'s may be either proper eigenvectors IE~I]) corresponding
to a discrete eigenvalue E~ and having normalization
(S.8a)
or generalized eigenvectors IE~ 1]) corresponding to the continuous eigen-

value E~ and having normalization
(S.8b)
Here peE) denotes the weight function (measure) with respect to which the
generalized eigenvectors of the continuous spectrum are normalized. 3 These
weight functions p(E) of the generalized energy eigenvectors IE) are a
convention which will be fixed for each particular case. For example, for the
energy eigenvectors of the beam system, Ip A) = lEBO, A), the weight
function p(E) is fixed by the normalization (S.7) and the usual convention for
the solid angle b-function b2 (Q - 0') given in (S.17) below. Since the IP A)'s
are also the generalized eigenvectors of the beam system's energy operator
(S.9)
the product vectors Ip A) ® lET 1]) are in fact the generalized eigenvectors
of the energy operator K = KB + K T . In (S.9) K~in(p) is the kinetic energy
operator for the projectile, the functional form of which depends on the
nature of the projectile; for a massive nonrelativistic projectile
K~n(p) = p2/2m, (S.10a)
while for a photon
(S.lOb)
where c denotes the speed of light. K~r('1°P) is the operator for the internal
energy of the projectile. The eigenvalues of K on the basis vectors (S.6) are
therefore
(S.11)
where
2
EB = L
2m + Eint(A)
B ,
11 Though K~t is the operator for the internal energy of the target (i.e., has no kinetic-energy
term) we shall not exclude the possibility that it also has a continuous spectrum, as may occur,
for example, in collision-induced scattering.
XIV.S Derivation of Formulas for the Scattering of a Beam off a Fixed Target 371
for a nonrelativistic projectile with internal energy levels E}ft(A), a structure-

less 12 nonrelativistic projectile, or a photon, respectively.
For the sake of definiteness we shall assume that the projectile is non-
relativistic and has no internal energy levels, so EB = p2/2m. For the other
cases in (5.11) or for the massive relativistic case EB = (p 2C2 + m 2c4 )1/2,
minor modifications will be necessary. We shall also use the abbreviated
notation
I
ET
=
Ef
f
I + p(E~) dE~ (5.12a)
and
<ET 1]IE~ r!') = bETET b~~, (5.12b)
for Equations (5.8a) and (5.8b). We may then express the product basis
vectors IP A) ® lET 1]) in terms of the total energy E and the spherical
coordinates of p:
lEa) = IEE TQA1]) = IpA) ® IE T 1])
= IEBQA) ® IE T 1]), (5.13)
where
p2
E = EB + ET = 2m + ET, (5.14a)
P = pO = p(sin e cos 4>, sin e sin 4>, cos e), (5.14b)

and
Q = (e, 4». (5.14c)
The b-function normalization of the product vectors IE a) follows from the
normalization (5.7) and (5.12a). We now determine the weight function
pee) in the normalization (2.9a) and (2.10). Since from (5.7) and (5.12a)
we have
= I
ETA~
f mJ2m(E - E T ) sin e de d4> dE, (5.15)
the basis vectors (5.13) must have-in terms of the quantum numbers E, E T ,
Q, A and 1]-the following normalization:
<EaIEa') = <EE T QA1]IE' E~Q' A.'1]')
1
J (j(E - E')(jETE'r(j2(Q - Q')(ju,(j~~, (5.16)
m 2m(E - E T )
12 More precisely, a projectile with only one energy level, which still may have different
polarization states.
where
b2 (Q - Q') = b(cos 0 - cos O')b( ¢ - ¢'). (S.17)
The weight function (measure) PaCe) is therefore fixed to be
PET(E) = p(E - E T ) = mJ2m(E - E T ) = mp, (S.18a)

and baa' of (2.9) is
(S.18b)
The mathematical relations that we need in the calculations are the
identities 13
f +OO
_ 00 dt e- iSI = 2nb(s) (S.19)
and
1 1
- - - - - = 2nib(s) (S.20)
s-iO s+iO
for the generalized function b(s).
We shall also need the relationship
(a+ IW(t)la'+) = (al WiR(t)la'), (S.21)
which can be derived immediately (Appendix XV.A) using the results
of Section XV.3 below. Equation (S.21) is plausible for the following reason:
Wet) is identical with WiR(t) in the distant past before the interaction
V becomes effective, and Ia +) is the same as Ia) in the absence of V, so
(S.21) holds at some time in the distant past. But the time developments of
Wet) and WiR(t) are generated by H and by K, respectively, which have the
same effect when applied to la+) and la), respectively; consequently, (S.21)
holds for all times t, not just in the distant past.
We now start the derivation of explicit formulas for the cross section.14
Inserting (2.18) into (5.3), one obtains
-i t ~ t+: elt e-i(Ea-Ea'JI<bl Vla+)<a'+ I Vlb)<a+ I Wla'+)
x (Ea. - ~b - iO - Ea - !b + iO)
= f +OO
-00 dtvoC1~(XAIW~(t)lxA). (5.22)
The right-hand side may be written by inserting the basis vectors IpA),
using the three-dimensional version
<xAlpA') = (2n)-3/2 ei P 'xb;.;.' (S.23)
13 Gel'fand and Shiloy (Vol. 1, pp. 168, 94).
14 The result of this tedious calculation is Equation (5.38).
of (11.7.51), the time development (5.5b) of WW(t), and the identity (5.19):
RHS of (5.22) = i f:: dt VoO" ~f d 3p dV<x AlpA)
x (pAle-iKBtWWe+iKBtlp' A)<P' AlxA)
= f+ dt Vo 0" L fd 3p d3p'(2n)- 3ei(p-p')'xe -i(EB-Es)t

00
-00 A
x <p AI WWlp' A)
= (2n)-2v o O" L fd 3p d 3p' ei(P-P')'xc5(E B - E~)

U'
x <p AI WWlp' A')b u
= (2n)- 2voO" L fp(EB)dEBP(E~)dE~dildil'

U'
where
p = po. = j2mE Bn., p' = pn' = j2mE~n.', and pee) = mj2mE

The calculation of the left-hand side of (5.22) goes as follows: We use (5.19)
to do the t-integration, (5.20) to replace the factor in parentheses, and (5.21)
to express things in terms of the state win(t), which may be broken up by
use of (5.4). The basis system { Ia)} used in the calculation is that consisting
of the generalized eigenvectors IEETilAI]) = IEBilA)® IETI]) with
E = Ea. So
baa'
x (a IWin Ia')2nic5(Ea - Eb )
= (2n)2 L L fp(E - E T) dE pee' - E'r) dE' dil dil'

b ET).~
E'r).'~'
x c5(E - E')c5(E - Eh)<bIVIEETilAI]+)

x <E'E'ril'A'I]'+lVlb)
x (EETilAI]IWinIE' E'ril' ,1'1]')
= (2n)2 L L fp(EB) dEB p(E~) dE~ dil dil'

b ET).~
E'r).'~'
X (j(EB + ET - E~ - E'r)c5(E B + ET - Eb )
x <bIVIEETilAI]+)<E' E'ril' A'I]'+lVlb)
x (EB il AI WW I E~ il' ,1') <ET I] I w!;' I E'r 1]') (5.25)
where
E = EB + Ey and E' = EB + E'y.

From (5.24) and (5.25) we then have
°
= (2n)2(RHS of (5.22) - LHS of (5.22))
- (2n)4 LL b(EB + Ey - EB - E'y)b(EB + Ey - Eh )

b ET~
E'r,1'
x <bIVIEEy~LlIJ+)<E' E'yQ' A'1J'+lVlb)<EYIJIW¥'IE'yIJ')}-

(5.26)
So far we have not specified any properties of the state W}f, but we know
already that the direction of the incident momentum is usually fixed in these
scattering experiments to be along the z-axis of Figure 3.1; i.e., 0. is fixed
to be 0. 0 = (0,0). The other properties of the incident beam-i.e. its polar-
ization and its energy distribution -can vary from experiment to experiment.
We shall continue the calculation under the additional assumption that
the beam is completely unpolarized, in which case the density matrix is
given by:
(5.27)
where g is the number of different polarization states A.. The reduced matrix
element <EB 0.11 W}fIIE B 0.') describes the momentum distribution of the
incident beam. Were the beam completely polarized with some definite
value ..1.0 for A., one would have to use
(5.28)
instead of (5.27) and would have to replace l/g by LA LA b

AAD in the calcula-
tions below.
As in the typical scattering experiment, the incident beam is always pre-
pared so that it has a well-defined (momentum) direction 0. 0 and the density
matrix of w}f has the property 15
for any smooth function F(n, 0.'). The doubly reduced matrix element
<EBlllw}fIIIE~> = f dndn'<EBnIIWk"IIE~n'>
= f dn<EB0.11 W}fIIE~ 0.> (5.29b)
of the state w}f with well-defined direction no describes the energy distribu-
tion in the beam, which will be discussed in more detail below. The normal-
ization of W}f requires that
1= ~ fp(EB) dEBdn<EBnAI w}fIEBnA>
= f P(EB) dEB dn<EB0.11 w}fIIEB0.>
= f p(EB) dEB<EBIII Wk"IIIEB>' (5.29c)
For an unpolarized beam (5.27) with well-defined direction (5.29a),

Equation (5.26) becomes
o = fp(EB)dEBP(E~)dE~<EBIIIW}fIIIE~>{voO"b(EB - E~)ei(p-p')no'x
(2n)4
- --II I b(E B + ET - E~ - E'r)b(E B + ET - Eb )
9 b.le ET~
E'r~'
x <bIVIE ET no A11+> <E' E'r no A11'+ IVlb> <ET 111 W~IE'r 11'>}-
15 One can convince oneself that (5.29a) is the continuous-spectrum analogue of (5.28) by
calculating for an arbitrary F,U.
H' H'
Intuitively one may want to write as the direct analogue of (5.28)
which, however, would be mathematically incorrect. Equation (5.29a) is an idealization; ac-

cording to the discussions in Section II.S every beam must have a finite spread in momentum.
Cf. also the discussion following (5.35) below.
As p = J2mEB' the exponential under the integral contributes only unity

because of the b(EB - EB)' Hence
o = fp(EB)dEBP(EB)dEB<EBIIIW~IIIEB){VOlTb(EB - E~)
(2n)4 , ,
- -- L L L b(EB + ET - EB - ET)b(EB + ET - Eb)
9 b.l ET'I
E'r'l'
x <bIVIE ET no A1]+) <E' E~ no AI]'+ IVlb) <ET I] I W~ IE~ I]')}.

(5.30)
Recall that we have left open the possibility that the target has continuous
as well as discrete energy levels E T ; although we have written (5.30) as if the
ET were discrete, it is also valid in the case (5.12) where there is a continuous
part to the spectrum of K T • In most experiments (although not for processes
like collision-induced scattering) the state of the target system is in a mixture
of discrete energy eigenstates (stationary states) before interacting with the
beam:
[K, WF] = [KT' WF] = o. (5.31)
Thus
WF = L <I]IIW!f(E~)III]')IE~I])<E~I]'1 (5.3]')
E~'1'1'
or
<E~ I] I W~IE* 1]') = bE~E¥'<1]11 W~(E~)III]')· (5.32)

Then (5.30) becomes
0= f P(EB) dEB P(EB) dEB(EBIII WknIIIE B)
, (2n)4 ,
x { volTb(EB - E B) - - - L L L b(EB - d
EB)b(E B + ET - E b )
9 b.l E~'1'1'
X <b IVI E E~ no A 1]+) <E' E~ no A1]'+ IVI b) <1]11 W~(E~)III]') }

= f P(EB) dEBP(EB)<EBIII W~IIIEB)
X {VOlT - (2n)4 L L L b(EB + E~ - Eb )

9 b), Ef}'1'1'
X <b IVI E E~ no A 1]+) (E E~ no A1]'+ IVI b) <I] I W~(E~)III]')}, (5.33)
where E = EB + E~ and E' = EB + E~. Before we can proceed with our

calculation we have to specify the energy distribution of the beam. Taking
the expectation value of the beam energy KB in the beam state W~ of (5.27)
and (5.29),
Tr(KBWkn) = ~ fp(EB)dEBdn(EBnAlwtKBIEBnA)
= f P(EB) dEB dn EB(E Bnil wtliEB n)
= f P(EB) dEB EB(EBIII w1nIIIEB), (5.34)

we see that
(5.35)
represents the probability distribution for obtaining the value EB when KB
is measured.
In the usual setup of scattering experiments the energy distribution is
peaked around some value E BO , which is under the control of the experi-
menter. Then F(EB - E BO ) is one of the functions F(x) depicted in Figure
11.8.1. It describes the energy resolution of the apparatus which prepares the
beam. The better the energy resolution, the narrower is the peak of
F(EB - EBO)' In the ideal but unphysical limit case of an exact beam energy,
F(EB - EBO ) will go into c5(E B - EBo).
Even in the realistic case of a finite energy resolution one can use as the
energy distribution function F(EB - E BO ) the generalized function
c5(EB - EBO ) if the T-matrix elements (bIVIEE~noAIJ+) are slowly
varying functions of E B • Let F be different from zero only in an interval of
width 211E around E BO -describing a beam with a small energy spread
E BO - I1E < EB < EBO + I1E-then the function F(EB - EBo ) will act like
c5(EB - EBO ) in an integral
f dEBF(E B - EBO)g(E B)
if the function g(E B) varies slowly over the range EBO - I1EB < EB <
EBO + I1E B. In (5.33) this will be the case if the matrix elements
(bIVIEE~noAIJ), where E=EB+EL vary slowly as functions of EB
over the interval 211E.16
16 Thus if the experimental resolution can be made narrower than the interval over which the
T-matrix varies significantly, the energy distribution of the state can be described by a generalized
function and the state by a generalized eigenvector whose energy wave function is a generalized
function. This generalized eigenvector is given by
lEo + iO) = limIT2~--

,-0 p (Eo)
f pCE') dE'IE') E, -J0r.
Eo +
. IE
and has as energy wave function the generalized function
<EIEo +. 1.
10) = -1-12~- lim
J0r. ..
P (Eo) ,-0 E - Eo - IE
It does not describe a physically preparable state and is not an element of 4> (cf. Section 11.8).
This is not always true. Near some values of E BO the matrix elements may
change very quickly; such a value E BO is called a resonance.
We shall discuss the effect of the finite resolution below and continue the
calculation here under the assumption that the beam is ideally mono-
chromatic, i.e., that
(5.36)
In this case the EB integration in (5.33) can be performed to get
(2n)4
0= P(EBO) { voa - - -" 1... "
1... "1... (j(EBO + ETd - Eb)<bl VIE ETnO
d 1 +
11.1] >
g b A E4~~'
x <E E~ no A1'1'+ IVlb> <1]11 W¥'(E~)III]'>}. (5.37)
where E = E BO + E~, Solving for a and reverting to rectangular coordinates

Po = J2mEBOnO' we at long last have our basic cross-section formula:
a(A +- Po) = (2n)4li L

2
2: L <Po AE~ IJ'+ IV Ib>
Vo b A EI}~~'
<IJII W~(E~)IIIJ'>(j(EBO + E~ - Eb),

X (5.38)
where EBO = p~/2m, Vo = Po 1m, and where we have introduced the notation
2: =!g).L
A
for the averaging over the polarizations in the initial state. We have also
restored the Ii's in order to express this important result in the usual units,
Equation (5,38) is the cross section for the scattering of an un polarized
beam with a well-defined momentum Po off a fixed target in a mixture W~r:
of discrete energy eigenstates and into any final configuration whose quan-
tum numbers b = (Eb' b) appear in the summation Lb' [To get the differen-
tial or partial cross section da(b +- Po) for scattering into the "state" Ab =
Ib> varies considerably when EB = E - E~ varies within
the interval AEB (this might happen in the neighborhood of a resonance
unless AEB is much smaller than the width of the resonance, as will be dis-
cussed in Chapter XVIII), Then the substitution (5.36) is not possible, and
consequently one cannot calculate a(A +- Po) as in (5.37). The quantity one
can calculate then from (5.33) is 17
J P(EB) dEB F(EB - EBO)va(E B) == (pva) * F, (5.39a)
17If the spread in velocity is not negligible, then the incident probability per unit area is
Jdz <xl W:;'(t) I x) = Jdt <xl 1"W:;'(t) I x) instead J
of dt vo<xl W:;'(t)lx), where 1" is the
velocity operator. This results in a replacement of Vo by V(EB) in (5.33).
This, according to (5.33), is then equal to
(pva) *F = (2n)4 fp(E B) dEBF(E B - EBO)LL I <bl VlpAE}1J+>

b ;. Ef}~~'
X <pAE~1J'+IVlb><1JIIW~(E~)II1J')c5(EB + E~ - Eb). (5,39b)

As P(EB) and V(EB) are slowly varying functions of EB compared to the
rapidly varying functions F(EB - EBO ) and l<bJVIEE~noAf/+>12, one can
take them out of the integrals on the left-hand side of (5.39a) and on the right-
hand side of (5.39b) and replace them by their average values, p(E BO ) and
V(EBO)' Then one obtains the cross section formula for coarse-resolution
experiments:
a *F = f dEBF(EB - EBO)a(EB) = (2n)4/i 2 v(Lo) fdEBF(EB - EBO)
x LL I c5(EB + E} - Eb)<bl Vip A EQ.1J+)

b ;. Ef}~~'
(5.40)
We continue now with the expression (5.38) for the idealized energy resolu-
tion, but will return to the case (5.40) of a limited energy resolution in Section
XVIII.8.
The cross section a(A +-- Po) is often expressed in terms of the T -matrix.
If the interaction Hamiltonian V is given, the T-matrix may be defined by
(5.41)
This does not fully define a transition operator T, because not all matrix
elements <E al TIE' a') of T are defined by (5.41), but only those matrix
elements" on the energy shell" E = E'. The matrix elements <E aITIE' a')
that are "off the energy shell," i.e., those for which E i= E', may be defined
in various ways; and one gets differing transition operators depending on
how these off-the-energy-shell matrix elements are taken. One such operator
is T + , defined by
(5.42)
Obviously the restriction of (5.42) to the energy shell gives the quantities
<EaITIE a') of (5.41). Another transition operator T- is defined by
(5.43)
Although not obviously, this too agrees with (5.41) when restricted to the
energy shell. Any useful definition of a transition operator T£ must agree with
(5.41) when on the energy shell:
<E al PIE a') = <E al TIE a'). (5.44)

Because of the presence of the factor c5(EBO + ET - Eb)
(which expresses
energy conservation) in (5.38), it is only the on-the-energy-shell T-matrix
elements that will contribute, a situation one always encounters when
dealing with physically observable quantities. Since all transition operators

agree on the energy shell, it makes no difference which T' is used.
In the S-matrix approach mentioned in the previous section, it is not the
existence of an interaction Hamiltonian V that is assumed, but rather that
of the T -matrix (E a ITIE a'), which is considered the fundamental quantity.
The expression of the cross section formula in terms of the T -matrix,
(2n)4Ji2 -
a(A +- Po) =
Vo
L L EIJ.~~'
b;'
L (j(EBO + E~ - Eb )
x (bl T IPoA E~l1)(Po A E~l1'l T I b)(1111 W~(E~)llll') (S.38')

may therefore be considered valid in either approach.ls If a V exists, then
(E hiT IE a') is defined in terms of (S.41) and
(E hi TIE a') = L (E hiE a)(E al TIE a'); (S.4S)
a
otherwise it is the T matrix (E a ITIE a') that is fundamental and which
cannot be further determined.
In order to calculate the cross section from (S.38) or from (S.38'), one has
to know both V and Ipo AE Tll+) or the matrix (E hi Tlpo AEP), respec-
tively. Even if V is considered to be known one still requires knowledge of the
Ipo AE~ 11+) = lEo E~Qo A11+) == IEoa+) (Eo = pU2m + E~), which are
solutions of the Lippmann-Schwinger integral equation. In principle the
IEo a+ ) may be obtained by an iteration process similar to the one described
in Section VIII. I. The approximation that one obtains for the lowest order of
the iteration process, i.e., the approximation
(bl T+ lEo a) = (blVlEo a+) ~ (blVlEo a), (5.46)
is called the Born approximation, and is usually sufficient in the case of a high
initial beam energy E BO and a weak interaction V.
Equation (S.38) [or (S.38')] gives the cross section for the rather general
situation in which the target state is left unspecified except for the require-
ment that it be a mixture of discrete energy eigenstates. We shall now give
the cross section for more specific target states. We first assume that W T is
diagonal in some of the quantum numbers fl of 11 = (fl, fi) and that no
measurement has been made with respect to the other quantum numbers fi.
For example, if the target consists of hydrogen or alkali atoms, the internal
quantum numbers 11 are 11 = (n,j,j3' n), and ET = ET(n,j, n) is a function of
n,j, and n. Usually the atom is in a specified mixture of energy and angular-
momentum and parity eigenstates, while the z-component of angular
momentum has not been measured. Then
. 1 .
(n j h nil W!F(E T) II n' j' j3 n') = 2j + 1 (j"", (j hj, (jjj' W!F(ET(n, j, n»(jnn'·
18 Strictly speaking, the matrix elements of an operator are the numbers that result from
placing the operator between two vectors from the same basis. Since the Sand Il may refer to
eigenvalues of different operators, (E S ITI E a) need not be a matrix element, but it is a linear
combination of matrix elements. For convenience, however, we shall hereafter refer to
(E SI TIE Il) as a T-matrix element.
Similarly, in the general situation
. d 1 . d
<~~II WT'(ET)IIiJ'iJ') = - c5ijW c5 iiii' WT'(E T, ij), (5.47)
gT
where gT is the number of values i'j can take on. The cross section obtained in
the case (5.47) is
(2n)4h 2 m - -
a(A ~ Po, ij) = ILL L I<b IT IPo A E'} ij i'j) 12
Po b)' E~ii ij
x W~(E~, ij)c5(EBO + E~ - Eb), (5.48)

where we use the notation L). L).
= (ljg) and L~ = (ljgT) I~ (averaging).
The weight W(E'}ij) of each energy level may depend upon all the quantum
numbers ijE'} = (1111'/2 ... 1'/K' E'}) or only upon a subset of them, e.g., 1'/1, E~.
Very often it depends only upon E~: W(E'}ij) = Wee'}). For example, if
the target system is in thermal equilibrium, then W T is given by (11.4.50)
and W(E~) is given by a Gibbs distribution
e-En/kT e- En/kT
Ween) = L e En/kT =L (dim Yfn)e En/ kT
where dim Yfn is the dimension of the energy eigenspace with eigenvalue En.
A very special but rather common situation is the case in which the target
is in a pure energy eigenstate, e.g., the ground state of an atom specified by
the quantum numbers (EO, 1'/0):
<1'/11 W~(E'})II1'/') = c5~~oc5~o~,c5Ef}E~' (5.49)

or
W¥' = IE~ 1'/o)<E~ 1'/0 I· (5.49')

The cross section (5.38) for scattering of an unpolarized beam off such
targets then becomes
2
0)
a(A ~ POET1'/O = (2n)4h m "L, "L, I<bl Vlpo A ET0 1'/0)1
+ Z
c5(Eo - E b)
Po b)'
(2n)4h Zm
=
Po
I I
b)'
l<blTlpo A E~ 1'/0)1 2 c5(Eo - E b), (5.50)
where Eo = E BO + E~. In the Born approximation and for structureless

projectiles, this reduces to the well-known formula
o (2n)4hZm " Z
I<bl Vlpo ET 1'/0)1 c5(EBO + ET - Eb)' (5.51)
0 0
a(A ~ POET1'/O) = L,
Po b
Usually, the transition matrix is independent of the polarization A. Then

the averaging over the polarizations in (5.50) is trivial, and the quantum
numbers A could just as well be omitted. For observables that do not depend
Upon the polarization, an unpolarized beam can be treated like a beam of
structureless projectiles; e.g., for an unpolarized electron beam one can

neglect the spin of the electron.
The most common experiments are those in which the differential cross
section for scattering into a particular solid angle AQDo is measured by
placing a counter at a particular angle QDO to the incident direction. AYf is
then the subspace of states with momentum pointing in any of the directions
QDO ± AQDo· If the detector detects only those states within a certain energy
range, then AYf is further restricted.
So far we have not specified the basis system Ib> = IE b> in the space of
final states. In order to obtain the differential cross section for scattering into
a particular direction specified by the angles Q = (8, <p), one conveniently
chooses a basis system of generalized eigenvectors labeled by Q or by the
momentum k of the detected (scattered) particle. Thus we choose
Ib> = IE b> = Ik 0 01t 0 = lED QD 0 01t 0
=IEDtQD~O· (S.S2)
where E = ED + I' and k = kn D. ~ are the internal quantum numbers of the
detected particle (polarizations) and t, , are the Internal energy and other
internal quantum numbers of the target. Thus our description is general
enough to include the case where the detected particle may be of a different
kind than that of the incoming beam, and the particle left behind (post-
collision target) may be different from those the target originally consisted
of. For example, a photon y might be incident on an atom A and ionize it,
leaving behind an ion A + with an electron coming off: y + A --+ e + A +.
A simpler example for the reader to keep in mind is the elastic or inelastic
scattering of an electron or photon beam by an atom: e + A --+ e' + A *.
If the particles detected are nonrelativistic, then
ED = k2j2mD + E~t(~), (S.S3a)
while if the particles detected are photons,
ED = kc. (S.S3b)
We shall impose the same normalizations as in (S.7) and (S.8) on the
IkO and ItO:
(S.S4)
and
(S.SS)
Then in the case (S.S3a) we have by derivations similar to those of (S.lS)
and (S.16) that
L= L
b
f 15b(E) dE = fL
s<~
mDJ2m D(E - E~t - 1') dE dQD
== L f
b
h(E) dE dn D (S.S6)
s<1;
and
(Ed1D~(I E'('O~C(,>
= mr;lj2mD(E - El;'t(~) - ()b(E - E')b a ,b 2(OD - O~)b~~,b~~,.

(S.S7)
Equation (S.SO) is then written in detail as
where
EO = P6/2m B + E~ (initial energy)
k' = j2mD(E' - Et't(~/) - (') (momentum of the detected particle)
The ranges of the summation over ~/, (', (' and the integration over E', O~
depend on the nature of the detection apparatus specified by A. From (S.S8)
we see that the differential cross section per unit solid angle for scattering
in the direction OD is
da (OD f- PoE~'1o) = (2n)4/i 2mBmD LL fdElkl(E I, (', 0

dOD Po A ~'£'~'
X 1(E' (' OD ~' ('IT 1Po .It. E~ '10> 12b(Eo - E').
(S.S9)
We shall now assume that the projectile and the detected particle are the
same (mB = m D = m) and that it is structureless or that the polarization
quantum numbers .It. and ~ can be ignored. 19 '10 and, are then eigenvalues
of the same set of operators '1 0P • The internal energies E~ or (', which are
eigenvalues of the same operator K T , are assumed to be already determined
by the internal quantum numbers, i.e., KT = Kl'('1°P ) as is the usual con-
vention for targets with discrete energy spectrum. 20 Then the expression
for (S.S9) simplifies to
dd~ (0 f- Po'1o) = (2n)4/i 2 m2 L fdE' kl(E', Ob(Eo - E')

~, Po ('
x I(E ' 0,('ITIPo'10>1 2 , (S.60)
where
k' = j2m(E' - ET(O) and Eo = P6/2m + E T('1o).
If not only the initial total energy is fixed (by fixing '10 and the beam
momentum Po) but also the final total energy (by choosing a detector that
19 One could also imagine that the polarization A. is incorporated in the quantum numbers 1'/.
20 If the target is the hydrogen atom. then '1 = (n,j,j3' 11:) and E T(II) = E; given by (VI.5.12).
registers only a particular momentum k), and if also the final internal quan-
tum numbers are fixed (by triggering the detector only if the final internal
quantum numbers have the value 0, then the differential cross section is
obtained from (5.60) as
d(J (21lf1i 2 m2 k 2
dQ(EQ(+---Po1]o) = Po I<EQ(lTIPo1]o)l· (5.61 )
where
k = J2m(E - ET(O),
which must be equal to
2m(~! + ET(1]o) - ET(O)
because of the enery-conservation b-function b(Eo - E). For the case of

elastic scattering, i.e., when the internal quantum numbers are not changed
(1]0 = 0, one obtains the well-known expression
(5.62)
The generalized eigenvectors in (5.62) [and also in the preceding expres-

sions (5.61) and (5.60)] are normalized by (5.54) and (5.7) according to
<E Q I] 1E'Q'1]') = <k 1]lk' 1]') = bqq ,b 3 (k - k').
If one uses generalized eigenvectors with a different normalization, then
this expression changes correspondingly.
For a spherically symmetric interaction, the T-matrix (and therefore the
differential cross section) will not depend upon the angle ¢ around the in-
cident direction Po/Po, but only upon the angle (J between the incident
direction and the final direction!l = k/k (cf. Figure 3.1). (A derivation of
this statement will be given in Chapter XVI.) e is called the scattering angle.
In this case one often uses instead of the T -matrix element the elastic-
scattering amplitude T(po, (J) defined by
T(po, (J) = -li4n 2 m<EoQ1]0ITlpo1]0). (5.63)
The differential cross section (5.62) is then written as
d(J 2
dQ (Eo, (J) = 1T(po, (J)I . (5.64)
Problems
1. The quantity q = k - Po, where Po is the projectile momentum before scattering
and k is the projectile momentum after scattering, is called momentum transfer. (It
represents the momentum that is transferred to the projectile by the target.)
Problems 385
(a) Show that the T-matrix <k, 110 I T IPo, 110) in the Born approximation is a function
of the momentum transfer only and does not depend upon the momenta
Po and k separately if the interaction Hamiltonian V is a function of the projectile
position.
(b) Show that in the Born approximation the forward T -matrix element
is independent of scattering energy.

(c) Assume that the interaction Hamiltonian V is spherically symmetric, [V, LJ
= 0, and express the momentum transfer in the Born approximation of the
T-matrix in terms of scattering momentum and scattering angle (for nonrela-
tivistic particles). Show that at high energies (when the Born approximation is
e
usually good), the scattering falls off as the scattering angle increases from
o to n.
2. Let the generalized eigenvectors Ip, A) = IE, Q, A) of the energy operator
K = p2/2m be normalized according to
Show that this fixes the weight function peE) for the energy normalization
Here Q denotes the direction of the momentum
P= po. = p(sin e cos rjJ, sin e sin rjJ, cos e),

and the solid angle J-function is defined by
J2(Q - Q') = J(cos 0 - cos e') J(rjJ - rjJ').
Calculate peE).
3. Obtain the expression for the differential cross section (5.62) in terms of the T-matrix
<EQITIEQ')
defined with the generalized eigenvectors IE Q), which are normalized according to
<E QIE' Q') = beE - E'W(Q - Q'),
where the solid angle b-function b(Q - Q') is defined by
fdQf(Q)()2(Q - Qo) = f(Qo)
with dQ = sin e de drjJ.

4. The Yukawa interaction was introduced to describe nuclear interactions and
led to the prediction of the meson. It is described by the potential
and can also serve as a simple model for the screened Coulomb field of an atom.
(a) Calculate the scattering amplitude T and the differential cross section in the
Born approximation.
(b) The second-order term in the Born series of the scattering amplitude is given by
[cf. (XV.3.38)]
T(2)(k <- Po) = -4n 2m<k 1J01 T(2) IPolJo>

= -4n2m fdV <kl Vlp'><p'l VI.Po>
Epo - Ep' + zO
Calculate the forward-scattering amplitude, i.e., the scattering amplitude for
Po= k, up to the second order of the Born series, and find the values of the
parameters m, fJ., and g, and the energies for which the necessary condition for the
validity of the Born approximation, IT(2) I ~ IT (1) I, is fulfilled.
(c) Obtain the total cross section for the Yukawa potential in the Born approxi-
mation.
(d) The parameter ro = 1/fJ. in the Yukawa potential characterizes the range of the
interaction (V(r) is practically zero for r > 2ro) and 9 its strength. Assume that
ro = 10- 13 cm and 9 = 0.5, as would be typical if VCr) is to descri,be the strong
interaction of protons scattering on nuclei. Let the projectile have the mass
m = mproton and the scattering energy E = 100 MeV. Calculate the value of the
total cross section.
CHAPTER XV
Formal Scattering Theory and

Other Theoretical Considerations
Notions used in Chapter XIV for the derivation of the cross-section formula
will be further discussed in this chapter to provide a deeper understanding
of the material. In Section XY.1 the Lippmann-Schwinger equation is
discussed again. Sections XV.2 and XV.3 are on formal scattering theory. In
the Appendix a derivation of (XIV.2.9b), (XIV.2.17), and (XIV.S.21) is given,
using the material developed in this chapter.
XV.1 The Lippmann-Schwinger Equation
The Lippmann-Schwinger equation

1
+ (1.1 ±)
la±) = la)
Ea - K ± I'0 Vla±)
was introduced in Section VIII.2. To each generalized eigenvector Ia) of K
with eigenvalue Ea the Lippmann-Schwinger equation relates a generalized
eigenvector la+) or la-) of H = K + V with the same eigenvalue Ea.
(In Section VIII.2 it was assumed that the continuous parts of the spectra
of H and of K were identical and that Ea was an element of these continuous
spectra.) The generalized eigenvectors Ia) of K and Ia±) of H are labeled
by the energy Ea together with some additional labels a = (aI' a2,"" ak ):
la) = IEaa), ( 1.2a)
la±) = IEaa±). (1.2b)
387
388 XV Formal Scattering Theory and Other Theoretical Considerations
a
In Section VIII. 1 it was assumed that the were additional quantum numbers
for the K basis, i.e., that
(1.3)
was a complete system of commuting observables (c.s.c.o) with the sets of
eigenvalues (E, a l' a 2 , ... , ak). Suppose that
(1.4)
is also a c.s.c.o. Both the generalized eigenvectors la) of (1.3) and the proper
and generalized eigenvectors Ian) and Ia + ) of (1.4) are a basis of the space of
physical states.
The assumption that (1.4) is a c.s.c.o. was not used in the derivation of the
Lippmann-Schwinger equation and may be relaxed accordingly. Suppose
that
{K, B 1, B 2 , .•. , Bd (1.5)
is a c.s.c.o. but that {H, B 1 , B 2 , .•. , Bd is not, i.e., that
[Bi' H] = [Bi' V] # 0 (1.6)

for some or all of the B;'s (i = 1,2, ... , k). By repeating the arguments of
Section VIII.2, the basis vectors
(1.7)
of K may be related by way of the Lippmann-Schwinger equation to the

generalized eigenvectors
Ib±) == IEb(b)±) = IEbb) + 1 . VIEbCb)±) (\.8)

Eb - K ± 10
of H. However, because of (1.6), the vectors IEb(b) are not eigenvectors of all
the operators B i • In this case the b;'s are defined not by a c.s.c.o. containing
H, but only through the Lippmann-Schwinger equation and the c.s.c.o.
containing K. The parentheses around the label b are a temporary notation
used to indicate that the b = b l , b 2 , . . . , bk are labels only, whereas the Eb
are eigenvalues of the operator H.
As an example of these possibilities consider the scattering of structure less
particles off a fixed target. Let Q and P be the (canonically conjugate)
position and momentum operators for the projectile, and assume the inter-
action of the projectile with the fixed target is given by a spherically sym-
metric potential V = V(Q) [Q = (Q2)1/2]; the Hamiltonian for the system
is then H = K + V, where K = p2/2m is the kinetic-energy operator for
the projectile. Since [L i , V] = [Li' K] = 0, it follows that
( 1.9a)
and
(1.9b)
XV.1 The Lippmann-Schwinger Equation 389
are both c.s.c.o's and that we have the case (1.4). The IEIl 3 ) and the IEIlD
are corresponding generalized eigenvectors. {I EIl 3 )} is a basis for the entire
space of physical states, but {I Ellt)} is a basis for the" subspace of scattering
states" only. The projectiles in scattering experiments are usually prepared
as collimated beams going in a specific direction, and it is therefore more
convenient to use for the K-basis the generalized eigenvectors
Ib) = Ip) = IEplP2) = IE(J¢) (1.10)
of the momentum operators Pi' The C.S.C.o. of the vectors (1.10) is
{Pj,P 2 ,P 3 } or {K,P 1 ,P2 } or {K,directionsofP}; (1.11)
we then have the case (1.5) with (1.6), because
[Pi' H] = [Pi' V] =1= o.
The label P for the H eigenvectors
I(b)±) = I(p)±) (1.12)
does not then mean that the I(p)±) are eigenvectors of the Pi but only means
that the I(p)±) are connected to the momentum eigenvectors Ip) by the
Lippmann-Schwinger equation,
(E = p2 /2m). (1.13)
The assumption that (1.3) is a c.s.c.o. for a particular physical system can
only be justified by physical motivation. It is equivalent to the assumption
that the set
{IE a): (E, a) E spectrum (K, A 1,···, A k )} (1.14)
of generalized eigenvectors of (1.3) forms a basis for the space of physical
states :It' of the physical system. Although the set
{IE a)H: (E, a) E spectrum (H, A 1, ... , A k )} (1.15)
of generalized eigenvectors of ( 1.4) is also a basis for ,Y{; the sets
{I E a+): (E, a) E spectrum (K, A 1, ... , A k )} (1.16+)
and
{I E a-): (E, a) E spectrum (K, A1, ... , A k )} (1.16 - )
of generalized eigenvectors of (1.4), which are obtained through the
Lippmann-Schwinger equations (1.1 +) and (1.1-), are in general not a
complete basis system for :It'. In the case that (1.4) is not a c.s.c.o., the sets
{I E(a)±): (E, a) E spectrum (K, A 1, ... , A k )} (1.17 ±)
are in general also not a complete basis system. Usually the operator K has
only a continuous spectrum, but H has in addition to the continuous spec-
trum a discrete spectrum {En} as well. The proper eigenvectors lEna) with
energy eigenvalues En in the discrete spectrum of H describe pure physical

states corresponding to the bound states of the projectile-target system, and
span a subspace of bound states Yf bnd' The orthogonal complement of Ye bnd,
i.e., the subspace of all vectors of Yf which are orthogonal to Yfhnd , is called
the space of scattering states '~cat. Thus 1
(1.18)
The subspaces Yf + and Yf _ that are spanned by the sets (1.16 + ) and (1.16 - )
need not necessarily agree with Yf scat or with each other. However we will con-
sider only physical scattering systems, for which the subspaces Yf +, Ye _,
and Ye scat agree:
Yf + = Yf - = Yf scat (1.19)
The system for which (1.19) is fulfilled is said to be asymptotically complete. 2
Under very general assumptions we thus have the following situation:
The space of physical states .It is spanned by the set {I E a)} of generalized
eigenvectors of K, and is the direct sum Yf = Yt' bnd EB Yf scat of the space
Yfbnd spanned by the set {I Ena)} of proper eigenvectors of H and the space
J'lscat spanned by either of the sets {I Ea +)} or {I Ea -)} of generalized eigen-
vectors of H related to the IEfl) by the Lippmann-Schwinger equation
(1.1 f).
The Lippmann-Schwinger equation is an equation for the Ia ±) and can
be iterated as described in Section VIII.1. It can also be solved formally. For
any two invertable operators A and B the following easily established re-
lation holds:
1 1 1 1 1 1
- - - = -(B - A)- = -(B - A)-. (1.20)
A B A B B A
A special case of this is
1 1 1 1
E- H ± iO - E - K ± iO = E- H ± iO V E - K ± iO' (1.21 ±)
By applying this to V Ia±) and then using (1.1 ±) twice, we obtain the
"solution"
+ 1
Ia -) = Ia) + Ea _ H +
_
iO V Ia). (1.22 ±)
t This statement is not difficult to envisage if the discrete and continuous spectra of Hare
disjoint. It may happen, however, that some of the discrete eigenvalues coincide with values of
the continuous spectrum (e.g., the doubly excited states of the He atom; see Sections XL2 and
XL3). To such eigenvalues correspond both proper eigenvectors in Jfbnd and generalized eigen-
vectors in the set of generalized eigenvectors that spans ,If,,,,, , Such proper and generalized
eigenvectors are orthogonal, even though they belong to the same eigenvalue of H and the
same eigenvalues a = (a b a 2 ,· .. , ak) of {At, A z ",·, Ad.
2 If the interaction V is described by a potential V(Q) which falls off more rapidly than r- 3
at infinity, is less singular than r- 2 at the origin and is sufficiently smooth in between, then
asymptotic completeness can be proven.
The operator analogue of Equation (XIV.S.20) is
1
E - H + iO E - H _ iO = - 2nib(E - H), ( 1.23)
which together with (1.22 + ) and (1.22 - ) allows us to relate the H eigen-
vectors Ia + ) and Ia -) to each other:
(1.24)
The formal solution (1.22 ±) is not of much practical use, but it does serve
as a way of introducing the S-matrix, which we shall meet in Section XV.3.
XV.2 In-States and Out-States
In this section we discuss the significance of the labels + and - that appear
in the generalized eigenvectors Ea +) and Ea -) of the exact Hamiltonian
1 1
H = K + V. In these discussions we shall consider pure physical states as

described by state vectors; the results are then easily carried over to statistical
mixtures as described by statistical operators.
Let us first describe the fictitious case where there is no interaction between
beam and target. This means that the beam passes through the target without
being affected at all. In the simplest case of potential scattering this means
that the potential is turned off. The state vector <1>(t) that describes this
fictitious situation develops in time according to the free Hamiltonian K:
(<1> = <1>(0». (2.1 + )
It is called the free state vector, and the state it describes is called the free state.
To simplify the discussion we shall assume that <1> is an eigenvector corre-
sponding to the eigenvalues £1= (a l , a 2 , ... , ak) of those operators A l ,
A 2 , ... , Ak that together with K form a c.s.c.o. (We assume the eigenvalues of
A l , A z , ... , Ak to be discrete.) Sometimes we will write <1>(£1, t) instead of
<1>(t) in order to emphasize <1>(t)'s preparation as an eigenstate of the A;'s.
The probability distribution for obtaining the energy eigenvalue E when K
is measured in the state <1> is
(2.2+ )
Let us assume that the energy distribution is described by the function cjJ(E),
which is related to <E a1<1» by
cjJ(E) = 2nPa(E)<E £11<1», (2.3 +)
where PaCE) is the normalization function for the generalized eigenvectors

of K and the generalized eigenvectors of H as given by (XIV.2.9a) and
(XIY.2.9b). That is, we assume that the state <1>(t) has been prepared at some
time such that at t = 0, and therefore at any other time t, it has the probability
distribution
<E o I<1>(t» <<1>(t) IE 0) = <E ol<1»<<1>IE 0)

= (2npaCE»- 21 ¢(EW (2.4+ )
for the eigenvalue E of the energy operator K. <1> may then be writteu in
terms of the generalized eigenvectors lEo) of K as
<1> = <1>(0) = f Pa(E) dE <Eol<1»IEo) = 2~f dE¢(E)IEo).

(2.5 +)
The free state vector at any other time t is given by
(2.6+ )
Let us now consider the state vector
<1>+ = <1>+(0) = 21nfdE¢(E)IEO+) = fpaCE) dE <Eol<1»IEo+),
(2.7 +)
where IE 0+) are the generalized eigenvectors of H that are connected with
the generalized eigenvectors lEo) of K by (Ll +). The H energy distribution
of <1> + is the expectation value of the" operator" La' lEo' + ) <E 0' + I:
<<1>+I(~IEO'+)<EO'+I)I<1>+) = ~1<EO'+I<1>+W = I<Eol<1»1 2.
(2.S +)
[Here we have used both the expansion (2.7 + ) of <1>+ and the normalization
(XIY.2.9b) of the IE o+)'s.] Thus, the probability distribution for obtaining
the value E when H is measured in the state <1>+ is the same as the probability
distribution for obtaining the value E when the free energy K is measured in
the (fictitious) state <1> [or <1>(t)]. For any other time t, earlier or later, the
state that develops to or from the state under the influence of the interaction
is given by
<1>+(t) = e- iHt <1>+ = 2~ fdE e-iEtlE o+)¢(E). (2.9+ )
<1>+(t) is called the exact state vector. <1> +(t) and <1>(t) are states with the same
quantum numbers 0 and the same energy distribution ¢(E); their difference
is that <1> +(t) develops according to the exact Hamiltonian and describes
the development of an actual state, while <1>(t) develops according to the
free Hamiltonian and describes the development of the same state as if
there were no interaction.
The connection between <1>+(t) and <1>(t) is obtained if one inserts the
Lippmann-Schwinger equation (1.1 + ) into the integral of (2.9 + ):
<1>+(t) = ~ fdE e-iEt¢(E)IE a) + ~ fdE e-iEt¢(E) 1 VIE a+)

2n 2n E- K + iO
= <1>(t) 1 fdE e-iEt¢(E)

+ -2 1 .0 VIE a+) (2.10+ )
n E-K+r
The first term in (2.10+) is given by (2.6+); the calculation of the second
term is a purely mathematical task and is given below, after (2.20-). Using
the result of that calculation one obtains
(2.11 +)
where
_ie iKt if t > 0,
Gri(t) = -i8( +t)e- iKt = { 0 (2.12+ )
if t < o.
B(t) is the unit step function
I if t > 0,
{ (2.13)
B(t) = 0 if t < o.
Suppose we now take the limit t -.. - r:tJ in (2.11 + ). As the limit is taken,
we eventually get t < t' for any value of t', so that the integrand vanishes.
Thus
(2.14+ )
This means that if in the remote past when the interaction was not effective
a state had been prepared so as to have the energy distribution ¢(E) (and
the quantum numbers a), then this state would develop so that at time t it
is given by <1>+(t) of(2.9+). Thus <1> + (t) describes a state that develops from a
state <1>(t) prepared in the remote past when the interaction V was not
effective. The prepared state is called an in-state:
(2.15+ )
Describing the behavior of <1>+(t) in the distant future when the interaction
V ceases to be effective is another free state, the out-state:
In the present situation, where the exact state <1>+(t) is prescribed by its
behavior in the distant past as <1>in(t), the outstate is unknown and un-
controlled, except through our knowledge of V and our control over <1>in(t).
As will be discussed in the next section, the connection between the in-state
<1>in(t) and the out-state <1>0UI(t) = S<1>in(t) is the scattering operator S.
We shall now repeat the same arguments used above but using the general-
ized eigenvectors IE 6-) of H that appear in the Lippmann-Schwinger
equation
~ ~ 1 ~
IEb-) = IEb) +E_ K _ iO VIEb-) (2.17 -)
instead of the IE a+) appearing in (1.1 +). Here 6 = (bb b2 , .•. , bk ) are the
eigenvalues of a set of operators Bb B 2 , ••. , Bk, which are possibly different
from the operators AI, A 2 , ••. , A k , but which together with K form a c.s.c.o.
(The spectra of B 1, B 2 , ... , Bk are assumed for simplicity to be discrete.)
We thus consider the free state
'P(t)[= 'P(b, t)] = e-iKt'P = ~ IdE e- iEt !/I(E)IE b), (2.6- )

2n
where the function !/I(E) describes the energy distribution of 'P(t) as it will be
prepared at some time. !/I(E) is connected with the probability distribution
for the energy operator K by
(2.4- )
The function i5s(E) is a weight function for the normalization of the IE 6)'s,
just as PaCE) was a weight function for the normalization (XIV.2.9a) of the
IE a)'s. From its expansion (2.6 -), 'P(t) is obviously an eigenvector of
B I' B 2 , .•. , Bk corresponding to the eigenvalues b.
The exact state vector 'P - (t) corresponding to 'P(t) is defined to be
(2.9- )
where IE b -) are the generalized eigenvectors of H that are connected with

the generalized eigenvectors IEb) of K by (2.17-). Again 'P-(t) and 'P(t)
are states with the same quantum numbers b and the same energy distribu-
tion !/I(E), but 'P-(t) develops according to the exact Hamiltonian H, while
'P(t) develops according to the free Hamiltonian K. To obtain the connection
between 'P- (t) and 'P(t) we insert (2.17 - ) into (2.9 - ) and obtain
'P-(t) = 2~ IdE e-iEt!/l(E)IE 6) + 2~ IdE e-iEt!/l(E) E _ ; _ iO VIE 6-)
= 'P(t) 1 IdE e-iEt!/l(E)

+ -2 1 . VIE b-). (2.10- )
n E - K - lO
The result of a calculation similar to the calculation to be given for the second
term in (2.10 + ) gives
(2.11 - )
{+
where
G- ( )
o t
=
+I
'8( _ ) - iKt
teo
= ie - iKt
°
if t < 0,
'ft > .
I
(2.12- )
If we now take the limit t -> + X) in (2.11 - ), we obtain

'P- (t) -> 'P(t) as t -> + 00 (2.14- )
This means that if in the distant future a state is measured that has the energy
distribution Ij;(E) and quantum numbers h, then this state was given at time t
by 'P-(t) of (2.9-). Thus 'P-(t) describes a state that will develop into a
known state in the distant future when the interaction V is no longer effective.
This state will be called an out-state:
(2.15- )
Describing the behavior of 'P-(t) in the remote past before the interaction
was effective is another free state, the in-state:
(2.16- )
Now the exact state 'P- (t) is prescribed by what its behavior will be in the
distant future as 'Pout(h, t), and it is the in-state that is unknown and un-
controlled. The connection between the in- and out-states is again 'P0ut(t) =
S'Pin(t)-or, since it is our control over 'P0ut(t) that should be emphasized,
'Pin(t) = S-l'P0ut(t) = st'P0ut(t).
We have thus found the meaning of the labels + and -: +(a, t) describes
a state that in the remote past, before the interaction V became effective,
was prepared as <l>in(a, t) with well-defined quantum numbers a and a certain
energy distribution epeE). In the distant future it will again become a free
state, an out-state <l>0ut(t); however, this state is not simple, since it is deter-
mined not only by the preparation but also by the scattering process.
'P-(h, t) describes a state that in the distant future, after the interaction Vhas
ceased, will be given by the free state 'Pbut(t) with a simple energy distribution
Ij;(E) and well-determined values h for the other quantum numbers. In the
remote past 'P- (h, t) was also a free state 'Pin(t), but its properties must have
been more complicated. Since in scattering experiments it is the behavior of
the system in the distant past over which we exercise control, it is the + states
<1>+ (a, t) that are natural to use when describing scattering experiments.
Our results may be summarized as integral equations for the exact states
+(a, t) and 'P-(h, t) in terms of the controlled free states <l>in(a, t) and
'Pout(h, t):
+(a, t) = <l>in(a, t) + tOOoo dt' Gri(t - t')V+(a, t'), (2.18+)
'P-(h, t) = 'Pout(b, t) + f:oo dt' Go(t - t')V'P-(h, t'). (2.18-)
(There are also corresponding integral equations for <1>+ (a, t) in terms of
<l>0ut(a, t) and for 'P-(h, t) in terms of'Pin(h, t), but in view of the uncontrolled
nature of <l>0ut(a, t) and 'Pin(h, t), such equations have little meaning.)
The preceding results are easily extended by linearity to the case of a

mixture. The exact state that was described in the distant past by the statistical
operator
(2.19 + )
mn
IS
(2.20+ )
mn
and the exact state that will be described in the distant future by
(2.19- )
mn
IS
(2.20- )
mn
We shall now take up the purely mathematical task of showing that

Equations (2.11 ±) follow from (2.10 ±), using results from distribution
theory, in particular results concerning the Fourier transforms of generalized
functions.
[The Fourier transform g = F[g] of a function get) is defined by 3
(2.21)
The inverse Fourier transformation F- 1 is given by
(2.22)
Strictly speaking, (2.21) is the definition of a well-behaved function.

The Fourier tranform of a generalized function f, defined by the
linear functional f('!') = (f, '!'), is defined by (F[f], F[,!,]) =
2n(f, '!'). But as the rules for Fourier transforms are retained for
generalized functions, when interpreted properly one may ignore
this mathematical precision and use (2.21) and (2.22) also for
generalized functions. 4 In particular, one always has
(2.23)
3 Note that only in this mathematical insert do we use the mathematician's convention for
the Fourier transform which is used in Gel'fand and Shilov (1964). in order to facilitate com-
parisons.
4 Gel'fand and Shilov (1964), Vol. 2. Chapter III; Vol. I, Chapter II.
Translation in the space of Fourier-transformed functions is

obviously given by
g(E - E') = F E[e - iE'tg(t)]. (2.24)
If we apply (2.24) to the Fourier transforms
1
FE[B(±t)] = ±i E ± iO (2.25±)
of the generalized function B( + t) and B( - t), we get
1 = F [+-ie-iE'tB(+t)]. (2.26±)
E - E' ± iO E -
The convolution f * g of two functions f and g is defined by
(f * g)(t) = J:oo dt' f(t - t')g(t') (2.27)
and has the particularly simple Fourier transforms

F[f * g] = F[fJ . F[g], (2.28)
or equivalently,
(2.29)
With the above mathematical facts we shall show the equivalence
of the second terms in (2.10 + ) and (2.11 + ). If we take the scalar
product of <1>+ (t) as given by (2.10 + ) with an arbitrarily chosen K
basis vector Ia') = IE' a') [more precisely, if we consider the value
of the functional I<1>±(t» E<1>x x = <1> at the point <a'i E<1>X], we
obtain
<a'I<1>+(t» = <a'I<1>(t» 1 JdE e-iEt¢(E)

+ -2 1 '0
n E - E' + I
X <a'i VIE a+). (2.30+)

Equation (2.9 + ) implies
<a'l V I<1>+ (t» = ~ fdEe-iEt¢(E)<all VIE a+)

2n
= Ft- 1 [¢(E)<a'lvlEa+)], (2.31+)
or equivalently,
¢(E)<a'IVIE a+) = F E[<a'IVI<1>+(t»]. (2.32+)
The second term on the right-hand side of (2.30 + ) is now recognized
as the inverse Fourier transform of the product of two Fourier
5 This statement has to be qualified somewhat in the case of generalized functions. See
Gel'fand and Shilov (1964), Vol. 2, Chapter III.
transforms [cf. (2.26+) and (2.32+)]. By using (2.29) we may

rewrite (2.30 + ) as
<a'I<1>+(t» = <a'I<1>(t» + (-W( +t)e-iE't) * «a'IVI<1>+(t»)
= <a'I<1>(t» - i f:oo dt' 8( +(t - t'»e-iE'(t-t')
x <a'IVI<1>+(t'). (2.33+)
Since Ia') was an arbitrary K basis vector, we have shown
1<1>+ (t» = I<1>(t» + foo dt' G; (t

00 - t')V I<1>+ (t'», (2.11 + )
as we set out to do. The equivalence of the second terms in (2.10 - )

and (2.11- ) may be shown by a similar calculation.
The operators
G~(t) = =Fi8(±t)e- iKt
are seen by (2.26 ±) to be the inverse Fourier transforms of the re-
solvents
G~(E) = (E - K ± iO)-l (2.34)
of the operator K. The operator G(j(t) is known as the retarded
Green's function, and Ga(t) is known as the advanced Green's
function. (In physics the distinction between a function get) and its
Fourier transform gee) = F E[g(t)] is not always reflected by the
terminology, and so the resolvents G~(E) are sometimes also called
"Green's functions," as mentioned in Section VIII.l. Physicists
frequently blur the notation as well, and write Gg:.(E) for Gg:.(E);
the argument of the function is then used to tell whether the function
or its Fourier transform is meant.)]
The operators G~(t) are both free Green's functions, as opposed to the
exact Green's functions given by
ift < 0,
G+(t) = -W( +t)e- iHt = {O . -iHt
-Ie ift> 0,
(2.35+ )
if t < 0,
G-(t) = +W(_t)e-iHt = {;ie- iH1 (2.35 -)
if t > 0.
These latter Green's functions may be used to express formal solutions of
(2.18 +) and (2.18 -):
<1> +(ii, t) = <1>in(ii, t) f+oo

+ _ 00 dt' G +(t - t')V <1>in(ii, t'), (2.36+ )
'1'- (I), t) = ,¥OU'(b, t) + f+~' dt' G - (t - t') Vqtou'(b, t'). (2.36 - )

XV.3 The S-Operator and the M011er Wave Operators 399
The derivation of these solutions, which is left as an exercise, parallels the

derivations of (2.11±) [or equivalently of (2.18±)] except that where the
derivations of the latter use the Lippmann-Schwinger equation (1.1 + ) and
(2.17 -), the derivations of the former use the formal solution (1.22 + )
and the formal solution for the IE b- )'s corresponding to (1.22).
One final remark should be made: None of our arguments have depended
on the assumptions that <Din(t) is an eigenvector <Dined, t) of the operators
Ap A 2 , ••• , Ak and that 'f0ut is an eigenvector 'fout(b, t) of the operators
B 1, B 2 , ••• , B k • All equations for pure states (as opposed to statistical
mixtures) are therefore equally valid, by linearity, if we remove the restrictions
that <Din(t) and 'fout(t) must be eigenvectors of the operators A I, A 2 , ..• , Ak
and of the operators B 1 , B 2 , •.. , Bb respectively.
XV.3 The S-Operator and the M~ller Wave Operators
In a scattering experiment a state is prepared before the projectile and target

start interacting with each other, and a state is detected after they stop
interacting with each other. Thus in scattering experiments in-states are
transformed into out-states. The operator that describes this transformation
is called the scattering operator or S-operator. For every physical system
undergoing collisions we postulate the existence of a unitary operator S
that transforms in-states into out-states. The S-operator is so immediately
related to the directly observable quantities (like cross sections) that know-
ledge of it leads at once to the prediction of these quantities.
As remarked in Section XIV.4, the concept of the S-operator makes sense
even if a generator H of time development is not defined and the time-
development axiom (V in Chapter XII) does not hold. If H is not defined,
then neither are its generalized eigenvectors Ia ±) nor the exact states
<D+(t) and 'f-(t). However, the states <Din and 'Pout, which describe the pre-
pared and detected states, and the operator
(3.1)
which transforms the prepared (and hence controlled) in-state <Din into the
uncontrolled out-state <D0ut, are still valid concepts. If the time development
axiom does hold (and there has been no evidence to the contrary in non-
relativistic quantum physics), then the S-operator and its matrix, the S-
matrix
(3.2)
may be expressed in terms of more fundamental and less directly observable
quantities.
We shall now do this by giving a precise definition of S in terms of pre-
viously introduced quantities. The modulus squared of the matrix element
(3.3)
gives the probability of finding the state \fI-(lJ, t), which is observed after the
interaction V has ceased to be effective as the state \fIout(b, t), if the state of
the system is <l>+ (a, t), which was prepared before V became effective as the
state <l>in(a, t). That is to say, (3.3) describes the probability for a transition
from an initial configuration, described by the quantum numbers and the a
energy distribution 4>(E), into a final configuration, described by the quantum
numbers b and the energy distribution tjJ(E). We first develop an expression
for (3.3) in terms of the in-state <l>in and the out-state \flout. To do this we shall
use the formal solutions (2.36 + ) and (2.36 - ),
(3.4+ )
and
\fI- (t) = \fIout(t) + t+: dt' G - (t - t')V\fIout(t'), (3.4- )
of (2.18 + ) and (2.18 - ). Equations (2.6 ±) and (2.15 ±) imply

<l>in( t') = e - iK(t' - O<l>in( t) (3.5 +)
and
(3.5 -)
These expressions may be substituted into the integrands of (3.4+) and
(3.4-) to get
(3.6+ )
and
(3.6- )
where we have defined the M¢ller wave operators
n± == I + J +OO
-00 dt' G±(t - t')Ve-iK(t'-t)
= I +iJ +oo.
_(jJ dt ' G±(t - t')VG~ (t' - t)
+ J
+OO
= I i -00 dt" G±( -t")VG~(l"). (3.7 ±)
An immediate consequence of (3.6 ± ) is that the mixtures W + (t) of (2.20 + )

and W-(t) of (2.20-) are related to the mixtures Win(t) of (2.19+) and
wout(t) of (2.19 -) by
(3.8 +)
and by
(3.8 -)
XV.3 The S-Operator and the Moller Wave Operators 401
Use of (3.6 ±) in (3.3) gives

('II-OJ, t), +(a, t» = ('P0ut(b, t), n-tn+inca, t» (3.9)
This last equation motivates
S == n-tn+ (3.10)
as the definition of the scattering operator.
The last expression for n ± in (3.7 ±) shows n ± to be time-independent,
which implies that S is also time-independent. So is the matrix element (3.9),
for +(t) is given both by
+(t) = e-iHt<l>+ = e-iHtn+in(o)
and by
+(t) = n+in(t) = n+ e-iKt<l>in.
Since {in} spans the space of physical states, we conclude that

e-iHtn± = n±e- iKt . (3.11 ±)
[The upper equations follow from the stated argument; a similar argument
gives (3.11- ).] Equivalently, one has the so-called intertwining relations
Hn± = n± K. (3.12±)
It is then easily shown that
[S, K] = 0, (3.13)
a consequence of which is the time independence of (3.9):
('P-(b, t), +(a, t» = ('P0ut(b, 0), e+iKtSe-iKl<l>in(a, 0»
= ('P0ut(b, 0), Sin(a, 0». (3.14)
Equation (3.13), it should be noted, may be interpreted as a statement of

energy conservation between the in-state <l>in and the out-state \flout.
If we substitute (2.6 + ) and (2.9 + ) into the left-hand side and right-hand
side of (3.6 + ), respectively, we obtain
Since 4>(E) may be any well-behaved function, it follows that

n+IEa) = IEa+). (3.15+ )
Equations (2.6 - ), (2.9 - ), and (3.6 - ) may be used in a similar manner to

conclude
(3.15 - )
Since we have assumed the system is asymptotically complete [Equation

(1.19)J, we may conclude that the M011er operators Q+ and Q- map the
space of physical states £ onto the space of scattering states £scat:
(3.16)
In other words, the domain of definition of both Q+ and Q- is the entire
space £, but the range of Q+ and Q- is the (proper if bound states exist)
subspace £scat.
[An operator A is said to be isometric if it preserves the norm of

vectors, i.e., if
(tf;, tf;) = (Atf;, Atf;) = (tf;, AtAtf;) (3.17)
for all vectors tf;. An equivalent definition is that A satisfies the
condition
(3.18)
If A satisfies the stronger condition

AtA = 1 and AAt = 1, (3.19)
then A is unitary. A unitary operator A is defined on all of £ and

has all of £ as its range, i.e., maps £ onto £ in a one-to-one
fashion. A unitary operator is necessarily isometric, but not vice
versa.]
Because Q± £ = £scat is usually not all of £, we should not expect the

M0ller operators to be unitary; they are, however, isometric, a fact we shall
now show.
By inserting
<1>+(t') = e- iH (t'-t)<1>+(t)
[which follows from (2.9 + )J into (2.18 + ), we obtain
<1>in(t) = <1>+(t) - t+: dt' Gri(t - t')Ve- iH (t'-t)<1>+(t)
(I + i t+: dt' Gri(t - t')VG-(t' - t)) <1>+(t) (3.20)
f:
=
= (1 +i dt" Gri( -nVG-( +n) <1>+(t).
The second line follows from the definition (2.35-) of G-(t) and the fact
that Gri (t - t') is nonzero only for t - t' > O. From the definitions (2.12 ±)
and (2.35 ±) of the Green's functions and from the Hermiticity of K and H
it easily follows that
(3.21a ±)
and
(3.21b±)
Upon use of (3.21 ±) and the definition (3.7 +) of n+, Equation (3.20)
becomes
<1>in(t) = (I - J:co
i dt" G+( -t")VGo(t,,)r <1>+(t) = n+t<1>+(t). (3.22+)
But <1>+(t) = n+<1>in(t), so
<1>in(t) = n+ tn+<1>in(t).
We have not made use of any property of <1>in(t) except that it can be expanded
in terms of the generalized eigenvectors IE a>
of K, so <1>in(t) is an arbitrary
element of Yl', Consequently,
n+tn+ = I, (3.23 +)
i.e., n+ is isometric. A similar derivation gives
n-tn- = I. (3.23 -)
The failure of n ± to be unitary is now easily shown: By use of (3.15 ±)
and (XIV.2.11) one obtains
n±n±t = (~n±la><al)(~n±la'><a'lr
(3.24±)
aa' a
= IT scat = 1 - IT bnd
where
IT scat = L Ia ± ><a ± I (3.25a)
a
and
(3.25b)
n
are the projectors onto the subspaces of scattering states and bound states,
Yl'scat and Yl'bnd, respectively. Unless the unitary deficiency IT bnd is zero,
i.e., unless there are no bound states Ian), the operators n± can only be
isometric and not unitary.
The adjoint M011er operators n± t map Yl'scat back onto Yl'; more precisely,
they annihilate the Yl'bnd part of Yl' = Yl'bnd EEl Yl'scat and map the rest,
Yl'scat' back onto ;if:
n±tl an> = (~n±la'><a'lrlan> = ~la'><a'±lan> = 0, (3.26±)
or equivalently,
(3.26' ±)
and
(3.27 ±)
We are now in a position to show that S is unitary:
sst = (n-tn+)(n+tn-) = n-t(J - TIbnd)n- = I (3.28a)
and
(3.28b)
In writing Equation (3.28) we have used the definition (3.10) of S, the iso-
metricity (3.23 ±) of n±, and Equations (3.24 ±) and (3.26 ±); in order for
Equation (3.28) to make sense-indeed, for the definition (3.10) of S to make
sense-we had to assume domain n±t = range n H = n+ Yr, or, in other
words, asymptotic completeness in the form (3.16). In this way we have
proven that the S-operator defined by (3.10) and (3.7 ±) in terms of Vand K
is unitary.
We define the S-matrix to be the matrix {Saa'} of the scattering operator
with respect to the K basis:
(3.10')
The S-matrix is therefore the transformation matrix between the two basis
systems (1.16 + ) and (1.16 - ) of the space of scattering states Yrscat :
(3.10")
By using the relation (1.24) we may write Saa' either as

Saa' = [(a+j - 2ni(aj V<5(Ea - H)Jja'+)
= (a+ja'+) - 2ni<5(Ea - EaYa j Vja'+) (3.29)
or as
Saa' = (a-j(Ja'-) - 2ni<5(Ea' - H)Vja'»
= (a-ja'-) - 2ni<5(Ea' - Ea)(a-j Vja'). (3.30)
If we use (XIV.2.9b) (which we shall finally prove at the end of this section),
we see that the first terms on the right-hand sides of (3.29) and (3.30) are
equal. Thus
I.e.,
(3.31)
In Section XIV.S we introduced the on-the-energy-shell T-matrix
7;.AE) == (E aj VjE a'+) (3.32)
(which does not define an operator) and its off-the-energy-shell extension
T:a, = (ajT+ja') == (ajVja'+) = (EaajVjEa,a'+), (3.33)
which defines an operator T +. Equation (3.31) suggests a different extension

of the T-matrix, namely
T:a, = <aIT-la') == <a-lVIa') = <Eaa-IVIEa, a'), (3.34)
which defines the operator T-. {T';+;',} and {T:a,} are extensions of {T"AE)}
in the sense that
T';+;', = T:a, = T"AE) when Ea = Ea, = E. (3.35)
The S-matrix element Saa' may then be expressed as
Saa' = <a Ia') - 2nic5(Ea - Ea,)T"a,(E a)
= p(Ea)-l(jaa,(j(Ea - Ea,) - 2ni(j(Ea - Ea,)T"AE a). (3.36)
Note that it is the on-the-energy-shell T-matrix elements that contribute to
the S-matrix; matrix elements of S between generalized eigenvectors lEa a)
and IEa, a') vanish for Ea ¥ Ea,·
An integral equation for the T+ -matrix may be obtained by inserting the
Lippmann-Schwinger equation (Ll + ) into (3,33):
T:;', = <al V(la') + E _ 1 '0 vla'+»)

a' K +1
= <alVla') + L <alVl a") <a" IV I.a'+)
a" Ea, - Ea" + 10
= <al VIa') + I <al Vla")T,;';'.a'. (3.37)

a" Ea , - Ea " + 10
Equation (3.37) may be iterated by repeated substitution of the right-hand

side into itself:
T';+;', = <al Via') + I <alVla")<a"l ~Ia') + .... (3.38)
a" E a , - E a" + 10
The Born approximation
(3.39)
in which only the first term of the iterative solution (3.38) is kept, is usually
adequate in the case of high energies Ea = Ea, = E and a weak interaction V.
From the unitarity of the operator S it follows that the S-matrix is also
unitary:
I
an
Saa"S:'a" = L S:"aSa"a'
a"
= (jaa' (3.40)
or in more detail,
~ f dEa"Pa,.(Ea,,)<E a aiSIE a" a") <Ea, a'ISIE a" a")*
= ~ f dE a" Pa,.(Ea")<Ea,, a"l S IEa a)* <Ea" a"l S IEa, a')

= p(Ea)-l(jaa,(j(Ea - EaJ (3.40')
By inserting (3.40') into (3.36) we may express the unitarity of {Saa'} In

terms of the matrix elements Taa.(E):
Taa,(E) - Tj,u(E) = -2ni I
{i"
PrAE)Taa,,(E)Tta" (E). (3.41)
The unitarity condition as expressed by (3.41) is sometimes called the

generalized optical theorem.
In summary, we have obtained both the S-matrix and the T-matrix by
assuming the axiom V of continuous time development. We have expressed
the S-matrix and T-matrix in terms of quantities, like Vand Ia+), that depend
on the existence of the time-development generator H [cf. (3.32) and (3.36)].
The generalized eigenvectors IE a+) of H and the corresponding states
t»
I<l>+ (a, describe the situation at the time of collision; they are therefore
quantities that are never accessible to experimental determination. In
contrast, the generalized eigenvectors IE a) of K and the corresponding
t»
states I<l>in(a, t» and I<l>0ut(a, are experimentally accessible by means of
the initial preparation and final observation of a collision experiment. They,
together with the operator that describes the transition between them, are
indispensable in the theoretical description of a collision experiment;
whereas experimentally nonaccessible quantities like I<l> +(a, t» are theor-
etical tools and rest on additional theoretical assumptions (e.g., the axiom
V).
The theoretical description which is based only on the experimentally
accessible quantities is called S-matrix theory. The S-matrix and the on-
energy-shell T-matrix, which are connected by way of (3.36), are then the
fundamental physical quantities; furthermore, the only assumption entering
into (3.36) is the conservation of energy between the in-state and the out-state
([S, KJ = 0). Unitarity, which implies the condition (3.41) on the T-matrix,
is a separate fundamental assumption in S-matrix theory and is justified by
the requirement of conservation of overall probability. Another fundamental
assumption in S-matrix theory is the analyticity of the T-matrix as a function
of energy and other physical parameters. This is argued to be connected
with causality. (We shall discuss this connection in Section XVIII.4.) In
the conventional formulation based on the assumption of a continuous time
development, properties such as unitarity and analyticity (in a certain
domain) are derived from the properties of H and of V.
Symmetry considerations place the most significant conditions on the
S-matrix. In the conventional formulation symmetries are formulated as
transformation properties of Vand H under various symmetry transforma-
tions, and these transformation properties then lead to derived symmetry
properties of the S-matrix. (We shall discuss the consequences of rotational
symmetry in Chapter XVI.) In S-matrix theory the symmetries are formulated
directly in terms of symmetries of the S-matrix or in terms of transformation
properties of the T-operator.
If V and H are known, then the conventional approach to the description
of collision processes is more "fundamental," although based on more
assumptions. After Equation (3.37) has been solved for the T-matrix, one
XV.A Appendix 407
may predict experimental results. As mentioned in previous chapters, in

physics we most often have the reverse situation: One wants to conjecture
from experimental data (e.g., cross sections) the properties of the fundamental
observables and then use these properties to make predictions about other
experimental quantities. In this situation a formulation of the properties in
terms of the T-operator is just as valid as a formulation in terms of V, and
is often more practical.
XV.A Appendix
With the help of the isometricity (3.23 ±) of the M0ller operators, 0 ± to ± = I,

it is now trivial to prove two equalities that were used in Chapter XIV.
(The proof of isometricity did not depend upon these equalities.) Equation
(XIV.2.9b) follows from (3.15±) and isometricity:
<a±la'±) = C<aIO±t)(O±la'»
(A. 1)
To prove (Xly'5.21) we use (3.15 + ) and (3.8 + ) together with isometricity:
<a+ I W+ (t)1 a'+) = C<a I O+t)(O+ win(t)O+ t)(O+ I a'»
= <al(O+tO+)Win(t)I(O+tO+)la')
= <a I Win(t)1 a'); (A.2+ )
in a similar fashion one can prove
<b-I W-(t)W-) = <bl WOU!(t)lb'). (A.2- )
Lastly we show that the term II given by (XIV.2.17) vanishes. The key to
doing so is Equation (A.2 +). If we substitute (XIV.2.9b) and (A.2 +) at
t = 0 into (XIV.2.17), then
11= L L e-i(Ea-Ea'll<bl Vla)p(Eb)-lba'bb(E a, - Eb)<al Winla')

baa'
= J P(Eb) dEb ~ J p(Ea) dE a p(E a,) dE a, fr, e - i(Ea -Ea'll

X <EbbIVIEa£1)p(Eb)-lb"'bb(Ea' - Eb)<Ea£11 WinlE a , a')
= f P(Eb) dEb ~ J p(Ea) dE a ~ e-i(Ea-Eb1t<Eb blVlEa a)
a
x <Ea I winl Eb b)
= L L e-i(Ea-Eb11<b IVia) <a I winlb). (A.3)
b a
But the summation Lb is only over those values of b for which Ib) E AX,
and by hypothesis the initial state Win is so prepared that I A) ¢ A£ for all
states IA) that give a nonzero contribution (W A =f. 0) to
Win = LwAIA)<AI.
A
Therefore
for all b in the summation, so that

11=0. (AA)
The contribution of II both to the transition rate d<A)t/dt of (XIV.2.5)
and to the transition probability <A) 00 of (XIV.5.2) is then, of course, zero.
CHAPTER XVI
Elastic and Inelastic Scattering for

Spherically Symmetric Interactions
In this chapter we derive general results for spherically symmetric situations,

i.e., for situations when in addition to the free Hamiltonian K, the interaction
Hamiltonian Vand the transition operator Talso commute with the operator
of angular momentum. For the sake of simpliciiy we consider the spin-zero
case. In Section XVI.1, the partial-wave expansion is discussed and the partial
cross sections are derived. In Section XVI.2, the phase shift is introduced as
a consequence of the unitarity of the S-matrix. In Section XVI.3, a graphical
representation of the partial-wave amplitude, the Argand diagram, is
introduced.
XVI.I Partial-Wave Expansion
Many scattering problems are spherically symmetric. This means that both
H = K + Vand Vcommute with the (orbital) angular-momentum operators
L; or, if the scattering is described by a transition operator T, that T (and
hence the scattering operator S) commutes with the L;'s:
[V, LJ = 0, [T, LJ = 0. (1.1)
Then
(1.2a)
and
(1.2b)
409
410 XVI Elastic and Inelastic Scattering for Spherically Symmetric Interactions
where I]0P and KOP stand for the internal observables (their eigenvalues I] and K
being the internal quantum numbers), are both c.s.c.o.'s, and one may choose
their generalized eigenvectors
(l.3a)
and
IE, 113 K) (l.3b)
as basis vectors of the space of physical states.
The vectors
(l.3c)
which are obtained from (l.3a) by the Lippmann-Schwinger equation
°
(XV.l.1 +) or by (XV.l.22+) are in general not eigenvectors of (l.2b). How-
ever, if [V,I]°P] = and therefore [H,I]°P] = 0, then the collection of
operators KOP = K~P", K~P in (1.2b) can be chosen to be I]0P. Then the
generalized eigenvectors (l.3b)-which are determined only up to an arbi-
trary phase factor that can still depend upon the quantum numbers-can
be chosen to be identical with (l.3c). In general [V, I]0P] =I: 0, and then (l.3c)
are not eigenvectors of the c.s.c.o. (1.2b) and the meaning of the label I] is
defined through the Lippmann-Schwinger equation as explained in Section
XV.!.
Still another c.s.c.o. consists of the projectile momentum P (or the momen-
tum P in the center-of-mass system when the target is not fixed) and the
operators for the internal quantum numbers I]0P.
(1.4)
The corresponding basis consists of the generalized momentum eigenvectors
IP 1]). (1.5)
The additional quantum numbers I] = (1]11]2" 'I]k) may refer to the
internal structure of the target, of the projectile, or of both. For example,
if the target is a hydrogen atom, then H will contain the quantum numbers
n,j,j3' and n that characterize the state of the electron in the hydrogen atom
(cf. Chapters VI and IX). According to (1.2), [L i , I]0P] = 0; thus we exclude
the possibility that I] includes the quantum numbers of an internal angular
momentum. Then the total angular momentum is equal to the orbital
angular momentum L i , and we avoid the complication of coupling the
internal angular momentum with the orbital angular momentum to get the
total angular momentum. Most processes encountered in experiment involve
at least one internal angular momentum; however, to understand the princi-
pal problems of this chapter it is easier if one omits at the beginning the
inessential complications introduced by the coupling of angular momenta.
The simplest case we can discuss is the case of a structureless projectile
incident on a structureless target. Whether the target and/or projectile may be
considered to be structureless depends on the energies involved. If, for
XVI.l Partial-Wave Expansion 411
example, the target is a helium atom, then it may be considered structureless

for projectile energies less than the energy of the first excited state, and the
quantum numbers 17 are unnecessary. If the projectile energy is comparable
in size to the difference between the energy levels of the helium atom, then
the target's structure must be considered, and 17 will contain the quantum
numbers n,j,j3' 7C (singlet or triplet) of the helium atom.
We would like to express the matrix element I
(1.6)
of the transition operator T with respect to the momentum basis (1.5) in
terms of the matrix elements of T with respect to the angular-momentum
basis (1.3a). To do this we expand Ip 17) in terms of the angular-momentum
basis,
Ip17) = Lfp~(E)dEIEI1317)<EI1317IP17)' (1.7)

113
The determination of the transition coefficients <P 17IE 113 17) = <E 113 17lp 17)*
is a problem similar to the determination of the wave functions <xln 1m)
of the hydrogen atom, which was discussed in Section VII.3. Instead of
(VII.3.5) we have the analogue
1 0
<P 17 IL;lI/I) = - T£ijk Pj OPk <P (}p4J p 17 11/1), (1.8)
from which follow the analogues of (VII.3.6 3) and (VII.3.6'). Here (p (}p 4J p)
are the spherical coordinates of the momentum vector p. Instead of the
"radial equation" (VII3.7) or (VII.3.1) we have
<P (}p (J~ + Kin}E 113 17) = E<p (}p 4Jp 17 IE 113 17),
4J p 17 I (1.9)
which is a consequence of the (nonrelativistic) relation (cf. XIV.5.31)
K = p2 + Kint(17°P) (1.10)
2M(17°P)
between the energy operator K, the momentum P, and the internal energy
Kint, Although the point has not been raised previously, the mass of the
particle may depend upon the internal quantum numbers 17 of the system;
this dependence allows us to consider cases where the outgoing particle is
different from the incoming particle. From (1.9) and the analogues of
(VII.3.6') and ~n.3.63) it follows that the transition coefficients have the
form [cf. (VII.3.!!)]
( E - ;~~ - E~n) YIl3«(}p, 4Jp) (1.11)
I The meaning of putting the quantum numbers in parentheses has been explained in Section
XV.I.
The function f~(p) is determined (up to a phase factor) by the normalization

(XIV.2.9a) of the generalized K-eigenvectors IE 113 YJ>, for
p~(E) -1(j(E - E')(jll' (j13 13 = <E 113 IJ IE' I' I~ YJ >
= Jd 3p <E 113 YJlpYJ><PYJIE' I' I~ YJ>
= J p2 sin (}pdpd(}pdcPpf;(p)(j (E - 2:~ - E~nt)

x yn/(}p, cPp)f~(p)(j (E' - ;~~ - E~nt) Y;'l,«(}p, cPp)
= m~p~(E)1 Np~(E»12(j(E - E')(jll' (j13 1" (Ll2)
In this calculation use has been made of (VII.3.l6) and the integration with
respect to p has been evaluated in the following manner2 using the substitution
t = l /2m~ - E:,nt, p = pet) = J2m~(t + E~t):
J p2 1J,(pW(j(E
~
- L
2m~
- Eint)(j(E'
~
- L - Eint)d P
2m~ ~
= J m~ J2m~(t
dt - E~t)1 f~(p(t»12 (j(t - E)(j(E' - t)
= m~ J2m~(E - E~t)1 f~(p(E»12(j(E' - E). (Ll3)
For the positive square root in (Ll3), which represents the absolute value
of the momentum as a function of energy, the following notation has been
used:
p~(E) = J2m~(E - E~nt). (Ll4)
Equation (Ll2), together with
p~(E) = m~p~(E) (1.15)
of (XIV.5.lSa), gives
(Ll6)
so that with an appropriate choice of phase for the generalized eigenvectors
>,
IP YJ the transition coefficients (1.11) are given by
<p(}pcPpYJIE 113 YJ> = -

(p~(E»-l(j (E - 2:~ E~nt) Y;1/(}p,cP p). (1.17)
Equation (Ll) tells us that Tis a scalar operator; its matrix elements with
respect to the angular-momentum basis are then [using the Wigner-Eckart
theorem (V.3.6)]
<E 113 YJbl TIE' I' I~ YJA> = <I' I~ 0 011l3> <E I YJbll TilE' l' YJA>
= (j1l,(j13 1,<E I YJbll TilE' I YJA>' (US)
, To evaluate the integral one can also use the mathematical formalism for 6(f(x)) developed
in Gelfand and Shilov (1964). Vol. I. Section II.2.S.
XVI.! Partial-Wave Expansion 413
Use of (1.7), (1.17), and (1.18) in (1.6) gives
<Pb'1bl TI PA'1A) = tf dE dE' t5(E - :lb - Ert ) t5 (E' - 2~A - E~t)

x <E l'1bIITIIE' l'1A) YIl3(Qb)yti 3(QA)' (1.19)
where OA = PA/PA I = (sin 8A cos tP A, sin 8A sin tP A, cos 8A) is the direction
of the incident particle and fib = Pb/lpbl = (sin 8b cos tPb' sin 8b sin tPb'
cos 8b ) is the direction of the detected particle. If we perform the summation
Ll 3 by using the addition theorem (VII.3.26)
+1 21 + 1
13~-1 YII3(Qb)Yti 3(QA) = ~ Pl(fib' fi A) (1.20)
for spherical harmonics, then (1.19) becomes

1
<Pb '1bl TI PA '1A) = -4 L (21 + 1)P1(cos 8)<E b1'1bll TIIEA 1'1A), (1.21)
nl
where 8 is the angle between fiA and fib (cf. Figure XIV.3.1), and where 3
(1.22a)
(1.22b)
The reduced matrix element <Eb 1'1bll TIIEA 1'1A) depends upon the energy
and the internal quantum numbers '1b and '1A, but does not depend upon the
direction. As one would have expected for a spherically symmetric problem
in which the initial conditions distinguish only the direction OA' the T-matrix
element (1.21) does not depend upon the angle tP around the direction 0A.
Equation (1.21) proves the statement preceding and justifying the definition
(XIV.5.63). Written in terms of the scattering amplitude, (1.21) becomes, for
'1b = '1A and Eb = EA,
T(PA, 8) = L (21 + l)PzCcos 8)( -nm)h<E A 1'1AII TIIEA 1'1A) (1.21')
1
In the expressions for the cross section [e.g., (XIV.5.38), (XIV.5.50), and
(XIV.5.60)] and in any other physical quantity, the factor t5(Eb - E A)
appears as a consequence of energy conservation [cf. (XV.3.13)]. Con-
sequently only the reduced matrix elements
<'1bll T,(EA)II'1A) == <Eb = EA 1'1bll TIIEA 1'1A) (1.23)
on the energy shell, i.e., for Eb = E A, have physical meaning. Energy con-
servation for the nonrelativistic case is given by
PA2/2mA + E ~A
int -
-
E A = E b -- Pb2/2mb + Eint
~b· (1.24)
In nonrelativistic collision experiments one usually has mb = mA, but does
not have E~".:. = E~n: except in the case of elastic scattering. In the following
3 Note the change in notation: EA here denotes the total energy, whereas in Section XIV.5 EA
denoted the projectile (kinetic) energy.
we shall examine in detail the case in which the detected particle is the same
as the incident particle, so that mb = mA; first, however, we give the expres-
sions for the cross section in the general (spin less) case.
The differential cross section per unit solid angle for scattering from an
initial state with momentum PA and internal quantum numbers YfA into a
final state with internal quantum numbers Yfb is [cf. (XIV.S.60) or (XIV.S.61')
at the end of Section XIV.4]
(1.25)
By using (1.21) we may rewrite da~b/dQb in terms of the reduced matrix
elements (1.23):
da~b 2 mAmb 2 "\' 2
dn. = n -~h Pb(E A) L (21 1 + I)PzCcos 8)(Yfbll 1[(E A)III1A) 1 . (1.26)
Hb PA I
The dependence of the momentum Pb = Pb(E b, Yfb) of the detected particle on

Eb( =E A) and Yfb is
Pb(E A) = J2mb(EA - E~nbt), (1.27)
as determined by (1.24) or (1.14).
The differential cross section given by (1.26) may be expressed as a series
in the Legendre polynomials PI> but the result is complicated (this will be
done in Section XVIII.8, where it is applied to the phase shift analysis). We
therefore give here only the expression for the total cross section, which may
be expressed simply in terms of the reduced matrix elements. By using the
orthogonality property
f -1
+1
d~ PI(~)PI'(~) = 21
2
+ 1 (jll' (1.28)
of the Legendre polynomials one obtains
mAmb
= 4n 3 -~ h 2 Pb(E A) I (21 + 1)1 (l1bll1[(E A)III1A) 12 (1.29)
PA I
for the total cross section. This suggests that we define the Ith partial cross
section
ai b == 4n 3 mAmb h 2 Pb(E A)(21 + 1) 1 (Yfbll1[(E A)IIYfA)1 2 (1.30)

PA
and that we write the total cross section as
(1.31)
XVI.1 Partial-Wave Expansion 415
ai b and a qb in (1.30) and (1.31) are the partial and total cross sections,
respectively, for scattering from a state with internal quantum numbers
IJ = IJA into a state with internal quantum numbers IJ = IJb' Often the
detector registers not just those states with a well-defined value IJb of IJ
but all states with values of IJ within some range. This happens, for example,
if IJ are the internal quantum numbers of the target and the detector registers
scattered projectiles with any possible energy. The projectile may excite the
target into any internal state IJ by losing the corresponding amount of
kinetic energy, and the detector cannot distinguish between elastically
scattered projectiles and inelastically scattered projectiles. In such a case the
lth partial cross section is given by
al = L ai b = 4n 3 mAmb 1i2(21 + 1) L Pb(E A) I <lJbll7;(E A)IIIJA> 12 , (1.32)

qb PA qb
and the total cross section is given by
a = L
qb
aqb = L al = L aib.
1 %1
(1.33)
Instead of the reduced matrix elements <lJbll7;(E A)IIIJA> it is convenient

to use the "partial-wave amplitudes." They are commonly used in nonrela-
tivistic scattering theory (particularly in the case of elastic scattering), and
formulas written in terms of them can be immediately taken over to the
relativistic theory. One proceeds from the assumption that the initial state
is fixed by the internal quantum numbers IJA> which usually describes the
state of lowest energy. From the reduced matrix elements one defines the
partial-wave amplitudes by4
n;)EA) = nb(PA) == -nJmAmbli<lJbll7;(EA)IIIJA), (1.34)

where PA is given in terms of EA and IJA by (1.22a) and in terms of Pb, E~bt,
and E~~ by (1.24). No additional quantum numbers IJb are required to de-
scribe elastic scattering, for which IJb = IJA' and one simply writes 5
(1.35)
The partial-wave amplitude nb(PA) for the inelastic case (lJb #- IJA) is
often called the partial-wave reaction amplitude, in distinction to the partial-
wave amplitude 7;(PA) for elastic scattering. In terms of the partial-wave
amplitudes the differential cross section (1.26) is written
da qb P (E ) 1 12
dOb = bPA A ~ (21 + l)Pz(cos O)nb(PA) , (1.36)
4 We have restored here the factor h which had been set equal to unity in the preceding
expressions for the cross sections. In (1.21') the factor on the right-hand side is h. With h = 1 the
length units, em, are identical to the inverse of the momentum units, erg sec/em or eV/c, and the
cross section will have the units of inverse momentum squared instead of area.
S We shall suppress the label tiA designating the internal properties of the initial state, when-
ever this does not obfuscate the notation.
while the Ith partial cross section (1.30) is written
(J?b = 41/ b(E A) (21 + 1)1 n b(PA)1 2. (1.37)

PA
The Ith partial cross section (1.32) for elastic scattering is then
(J/(elastic) = (J?A = 4n(21 + 1)1 1!(PA) 12, (1.38)
and the Ith partial cross section for inelastic scattering is
4n
(Jlinelastic) = L (J?b =- (21 + 1) L Pb(EA)1 n (PA)1 2.
b (1.39)
~b*~A PA ~b*~A
The final momentum Pb(E A) that appears in the above equations may be
expressed in terms of the initial momentum and the difference between the
initial and final internal energies by
Pb(E A)2/2mb = p~/2mA - (E~nt - E~t). (1.40)
The S-matrix is given in terms of the (on-the-energy-shell) T-matrix by 6
<Eb 113 '1bISIEA l' I~ '1A) = <Eb 113 '1bIEA l' l~ '1A)
- 2ni6(Eb - EA)<E A 113 '1bl TIEA I' l~ '1A)' (1.41)
Spherical symmetry, which was initially stated by (Ll), is equivalently stated

by
[S, LJ = O. (1.42)
S is thus a scalar operator and by the Wigner-Eckart theorem has matrix

elements of the form
<Eb 113 '1bl S1EA l' I~ '1A) = 611'6/ 313 <Eb I '1bIISIIEA l'1A)' (1.43)
From energy conservation [Equation (XV.3.13)],
[S, K] = 0, (1.44)
it follows that
<Eb I '1bIISIIEA l'1A) = 6(Eb - EA)Pb(EA)-1/2PA(EA)-1/2<'1bIIS/(EA)II'1A)'
(1.45)
where p~(E) is given by (Ll4) and (1.15). The factoring off of the weight
functions Pb 1/2 and PA 1/2 is a convention made in the definition of
<'1bIISlEA)II'1A) [but not in definition of <'1bll71(E A)II'1A)] in order to simplify
later expressions (e.g., the unitarity condition in terms of the S- and T-
matrices, which then can be carried over directly to the relativistic case).
6 The S- and T-operators were discussed in more detail in Section XV.3, and (1.41) is identical
with (XV.3.36). If Chapter XV was skipped in the first reading, the reader may take (1.41) as the
definition of the S-matrix.
XVI.2 U nitarity and Phase Shifts 417
After substitution of (1.18) and (1.23) and of (1.43) and (1.45), the connection
(1.41) between the reduced matrix elements of Sand T becomes
<'1bII SI(E A)II'1A) = bqbqA - 2niJ Pb(EA)PA(EA)<'1bllll(EA)II'1A)' (1.46)
In terms of the partial-wave amplitudes this connection is given by
Sz{EA) == <'1AIISz{EA)II'1A) = 1 + 2iPAll(PA) (1.47)
for elastic scattering and by
S?b(EA) == <'1bII Sz{E A)II'1A) = 2iJPb(EA)PAT?b(PA)
for inelastic reaction processes.
XVI.2 U nitarity and Phase Shifts
In the previous section we expressed the cross section in terms of the transition
matrix and also in terms of the partial-wave amplitudes [Equations (1.25)
and (1.36)]. The only condition used was the assumption (1.1) of spherical
symmetry. This allowed us to write the cross section in terms of some known
functions of the scattering angle and the partial-wave amplitudes. For a
given scattering problem the partial-wave amplitudes can be determined
once the potential is known, in the same way that the scattering amplitude
can be determined from the potential, by solving the integral equation
(XV.3.37). There is, however, a general condition that permits us to obtain
some general properties of the scattering amplitude. This is the unitarity
of the S-operator, which is an expression of the conservation of overall
probability and which is a consequence (for the case that time development
is described by a Hamiltonian, as specified by the axiom V) of the Hermiticity
of the Hamiltonian.
The unitarity of the scattering operator [cf. (XV.3.28)r
sts = I (and sst = 1), (2.1)
is written in the angular-momentum basis as
I fPb(Eb) dEb<E 113 rylStlEb113 '1b)<Eb113 '1bI SIE A l' I~ '1A)

113~b
- ~~ 1 - •
= <E 113 rylE A l' I~ '1A) = PA(E A)- beE - EA)<'hl' (jf31,(jijqA' (2.2)
Using (1.43) and (1.45), this leads to the unitarity condition for the reduced
S-matrix elements:
(2.3)
For the special case ij = '1A Equation (2.3) may be written as

ISz{EA)1 2 + L: 1<'1bII Sz{E A)II'1A)1 2 = 1, (2.4)
7 For the case that one does not assume a Hamiltonian time development (2.1) is taken as
one of the basic postulates for the S·operator.

where Lnbextends over all possible values of I1b except I1b = I1A. Inserting
(1.47) and (l.48) into (2.4) leads to the unitarity condition expressed in
terms of the partial-wave amplitudes:
= PAl ll(PA) I2 + L Pb(EA) IT7 b (PA)1 2 • (2.5)

~b*~A
Equation (2.5) is the partial-wave amplitude form of the "generalized optical

theorem" (XV.3.41). The first term on the right-hand side corresponds to
elastic scattering; the second term corresponds to inelastic reaction processes.
For a particular collision experiment the values of I1b over which the sum
in (2.3) and (2.5) extends may be limited to those permitted by the initial
energy. If the kinetic energy ofthe projectile is lower than the energy necessary
to excite the target, i.e., if
2/2m A
PA < Eint _
~b
E int
~A
(for an "infinitely heavy" target that remains always at rest), then the pro-
jectile can only be scattered elastically, and all (l1bIIS,(EA)III1A) vanish for
I1b -# 11k One says that only the "elastic channel" is open. For instance, if
in e-H scattering the kinetic energy p2/2m of the incident electron is lower
than the energy difference between the ground state and the first excited
state, E(n = 2) - E(n = I), then only elastic scattering
e+H-+e+H
is energetically possible, so only the elastic channel is open. As the kinetic
energy of the incident beam is increased, one reaches a point above which
it is energetically possible to raise the target into the first excited state 11 1
with the projectile losing a corresponding amount of kinetic energy. This
point is the threshold of an inelastic process. The threshold momentum for
an infinitely heavy target is given by
2
Pthresh /2 mA -- E ~1
inl
-
int
E A·
Above this value for PA, both (I1AIIS,(E A)III1A) and (1111IS,(EA)III1A) are
different from zero. One says that an "inelastic channel" has opened up.
As the incident energy is further increased, other inelastic processes may
become possible, and an increasing number of reduced S-matrix elements
in (2.4) become nonzero.
There can be different kinds of inelastic channels. The simplest are the
excitation channels, in which the internal state of the projectile (target)
remains unchanged whereas the target (projectile) performs a transition
from the initial state into an excited state.I1A and I1b then refer to the different
internal quantum numbers of the target (projectile). (If the projectile is
also a system with internal structure and its internal quantum numbers are
contained in the set 11, then these quantum numbers will remain fixed.) In
XVI.2 U nitarity and Phase Shifts 419
the example of e-H scattering the excitation channels are the final states of
the process
e + H(n = 1) -+ e + H*(n 2 2).
The threshold energy for this process is 1E(n = 1) - E(n = 2)1. If the incident
energy is increased further, the target may ionize, i.e., the target may be
raised to the energy at which it breaks up into two particles; one then says
that the "ionization channel" has opened up. For e-H scattering the ioniza-
tion channel is the final state of the process
e + H(n = 1) -+ e + e + H+
and the ionization threshold is IE(n = I) - E(n -+ 00)1 = 13.6 eV.
The internal quantum numbers '1b, or at least a subset of them, now also
extend over a continuous set of values. In these cases it is often more practical
to use for the postcollision system a different set of quantum numbers than
for the precollision system. This becomes even more necessary for more
complicated collision processes like rearrangement collision, in which the
projectile is absorbed by the target and another particle is emitted. A simple
example of such a process in atomic physics is
}' + He -+ He + + e.
F or more complicated precollision systems the number of possible scattering
channels increases-e.g., in the collision of He atoms and protons the
following channels are possible:
H+ + He -+ H+ + He (elastic channel)
-+ H+ + He* (excitation channels)
-+ H + + He + + e (ionization channel)
-+ H + + He +* + e (ionization-excitation channels)
-+ H + He + (rearrangement channel)
-+ H* + He+ (rearrangement-excitation channels)
Many-channel processes also occur in nuclear physics and in particle

physics, and it is in these areas that many-channel problems become particu-
larly important. For instance in pion-proton (n-p) scattering one has (among
others) the following channels:
n +p-+n +p (elastic channel)
-+ n- + nO + p
-+'1+n
-+ KO +A
420 X VI Elastic and Inelastic Scattering for Spherically Symmetric Interactions
(1J and KO denote mesons heavier than n; n and A denote the neutron and the
"strange" baryon A, which is heavier than the proton and the neutron;
p* denotes various excited states of the proton.)
All scattering channels (i.e., the elastic channel and all inelastic channels)
are simultaneously open if the initial energy is high enough. These scattering
channels are various final states, differing in the internal quantum numbers
1Jb, and the sums in (2.4) and (2.5) must be extended over all channels that
are open, i.e., energetically possible. The interaction can, of course, be such
that an open channel, say 1Ji" does not contribute much to a particular
partial-wave amplitude, i.e., that 1<1JiiIIStCEA)II1JA)1 is small or zero even for
energies above the threshold for the channel1Jb' These are detailed properties
of the interaction (" the dynamics") which require further information.
We initially assume that all inelastic channels are closed. Equation (2.4)
then becomes
(2.6)
For elastic scattering it thus follows that StCE A) is a function of EA (or,
equivalently, of PA) of modulus 1 and consequently may be written
(2.7)
where (5,(E A) is a real-valued function defined up to an integral multiple of n
by (2.7). (5/(E A) is called the scattering phase shift for reasons we will discuss
later.
The elastic partial-wave amplitude that appears in (1.47) may also be
expressed in terms of the phase shift,
(2.8)
PA cot (5/(PA) - i'
which in turn allows us to express the lth partial elastic cross section of (1.38)
in terms of the phase shift:
o"t{elastic)
PA
= 4:
(21 + 1) sin 2 (5tCP A)' (2.9)
We would like to extend the phase-shift formalism to include the inelastic

or reaction channels. If the sum over the reaction channels in (2.4) is nonzero
(i.e., if EA is sufficiently high that some of the reaction channels are open),
then StCE A) can no longer be written as in (2.7) with real (5tCE A); but it may be
parametrized in the formS
StCE A) = 1JtCE A)e 2ib ,(E A ) (2.10)
8 The '1, that appears in (2.10) is not to be confused with our use of '1 to denote the internal
quantum numbers. We use it here because this notation of ii, and '1, is often used in particle
physics; in other areas of physics the notation may be different-e.g., in atomic physics '1, is
often the symbol for the phase shift denoted by ii, here.
XVI.2 Unitarity and Phase Shifts 421
with real blE A) and with real '1lEA) satisfying

o s '11(E A) s 1. (2.11)
That '11 satisfies (2.11) may be seen by inserting (2.10) into (2.4),
'1lEA)2 = 1 - L I<'1bII SI(E A)II'1A)1 2 , (2.12)
%*~A
and observing that
cannot be larger than 1. Equation (2.12) can be written in terms of the partial-
wave reaction amplitudes by using (1.48):
(2.13)
The partial-wave amplitude for the elastic channel may be expressed in

terms of blpA) and '1lPA) as [cf. (1.47)]
T _ SI(P A) - 1 _ '1lp A)e 2;Ih(PA) - 1
l PA) - 2· - --'-'---''-=-----:::-2-·- - -
Ip A IPA
'1lp A) sin 2c5 1(PA) . 1 - '11(P A) cos 2c5lpA)
= +I . (2.14)
2PA 2PA
From (1.38) and (2.14) it then follows that the lth partial elastic cross section is
o"zCelastic) = ~(2l + 1)(1 + '1lPA)2 - 2'1z{PA)cos2blpA)), (2.15)

PA
while from (1.39) and (2.13) we have
oiinelastic) = ~ (21 + 1)(1 - '11(PA)2) (2.16)

PA
for the lth partial inelastic cross section. The total lth partial cross section
is the sum of these:
2n
(Jl = (Jz(total) = 2" (2l + 1)(1 - '11(PA)cos2b1(PA)). (2.17)
PA
For '11 = 1 both (2.15) and (2.17) are identical with (2.9), which gives (JI in
the case that all inelastic channels are closed, and (2.16) vanishes. From this
and from (2.10) and (2.11) one sees that '11 is a quantitative characterization
of the amount of inelasticity involved in the scattering; '11 is therefore called
the inelasticity coefficient.
The total lth partial cross section may also be expressed in terms of the
lth elastic partial-wave amplitude by adding (1.38) and (1.39) and using the
generalized optical theorem (2.5) to get
4n
(Jl = -(2l + l)Im ll(PA). (2.18)
PA
If we sum (2.18) over alii and use the imaginary part at = 0 of the elastic- e
scattering amplitude (in the form obtained from (1.21') and (1.35)),
T(PA, e) == L(21 + I)PI(cose)TtCPA), (2.19)
I
we then obtain
4n 4n
a = L al = - I (21 + 1) 1m TtCPA) = - 1m T(PA, e = 0). (2.20)
I PA I PA
[Also used was the property PI(l) = 1.] Equation (2.20), which connects
the total cross section and the imaginary part of the forward-scattering
amplitude, is usually called the optical theorem.
XVI.3 Argand Diagrams
In order to familiarize ourselves with the properties of the scattering ampli-

tude and its behavior under certain conditions we consider a graphical
representation in which PA Tt(P A) is regarded as a vector in the complex
plane. These graphical representations are called Argand diagrams and play
a particularly important role in the description and detection of resonances.
We start from the expression (2.14) for the elastic partial wave amplitude,
which is written in the form
PA Tt(PA) - ~ = -~ '11(P A)e 2iIi1 (PA). (3.1)
F or the case of purely elastic scattering (e.g., in the region oflow momentum
where no inelastic channels are open) '11(P A) = 1, so PA Tt(P A) - i/2 is a
complex number that lies on a circle of radius 1/2. If 6tCp A) varies between
o ~ 61 ~ n (modulo n) when PA is varied over the range in which only elastic
scattering occurs, then all points of this circle are covered. In general, how-
ever, 6tCp A) will not vary over all of the interval 0 ~ 61 ~ n, and hence
PA Tt(p A) - i/2 will not vary over the full circle as PAis varied over that range
in which only elastic scattering is possible.
For elastic scattering the values of
(3.2)
lie on a circle with center at i/2 and radius 1/2. This is in agreement with the
optical theorem (2.5), according to which 1m Tt(p A) Z O. The circle around
i/2 with radius 1/2 is called the unitary circle.
If there is absorption of energy by inelastic processes, then
PA Tt(p A) = ~ - ~ '1 tCP A)e 2iIi1 (PA) (3.3)
and the value of PA Tt(p A) moves on a path inside the unitary circle as PAis
varied. Figure 3.1 shows the graphical representation (Argand diagram) of
XVI.3 Argand Diagrams 423
o +1
Figure 3.1 The elastic-scattering amplitude PAT/(p A) in the complex plane. (Argand
diagram.)
PA TlpA)' One calls the circle the" unitary circle" because, as a consequence
of the unitary condition, which is expressed by (3.3), it follows that the path
of PA 7;(p A) must lie inside this circle. The exact path followed by PA 7;(p A)
depends upon the particular interaction. The most important case of these
diagrams is the case in which the interaction between projectile and target
is resonant. We will consider these cases below in Section XVIII.8.
Argand diagrams may also be drawn for the partial-wave amplitude
T(b(PA) of an inelastic channel (lJb ¥- IJA)' One expresses the partial-wave
amplitude T7 b(PA) [or the reduced S-matrix element <lJbIISlEA)IIIJA) that
corresponds to it by (l.48)J in a form analogous to (2.10):9
Jpb(EA)PAnb(PA) = ;i<lJbII SI(E A)IIIJA) = ~ IJr(PA)e 2ibf (PA): (3.4)
Unitarity of S((2.3) or (2.4)) imposes the restrictions

0:::;; IJr(PA) :::;; 1) (3.5a)
and from (2.12)
IJI(P A)2 + L IJr(p A)2 = 1. (3.5b)
%*~A
In the Argand diagram, a == JP1E:)p~ T7 b (p A) then moves as a function

of PA on a path that lies inside the circle with center at the origin and radius
1/2 (cf. Figure 3.2). (The path would lie on the circle if there were neither
elastic scattering nor inelastic scattering into any channel other than the
one with quantum numbers 'lb')
" IJb , IjA denote the internal quantum numbers (channels).Ij~ denotes the inelasticity coefficient
for channel IJb as in Equation (3.4).
1m :.:
+1
~--------------~---1----------~~Re a
-1
-1
Figure 3.2 The amplitude IX = Jpb(EA)PA T7 b(PA) for scattering from channel '1A to an
inelastic channel '1b ' In this case the center of the circle is at the origin and the amplitude
is restricted to IIX I :$; ! .
Problems
1. (a) Show that the Born approximation violates unitarity.

(b) Where in the Argand diagram for elastic scattering can the Born approxima-
tion be considered a good approximation?
2. Obtain the inelastic and the elastic total cross section for scattering (of a structure·
less projectile) by a black disc (hard sphere) of radius a. A black disc is described by the
following properties: It is totally absorbing, i.e.,
(a) '11 = 0 (zero inelasticity coefficient), and
(b) only partial amplitudes with I :$; L contribute to the cross section where
L = PAa.
«a) specifies that the disc is black, (b) specifies that it has a well-defined edge.) Discuss
the result, in particular the value for the elastic total cross section.
CHAPTER XVII
Free and Exact Radial Wave

Functions
The present chapter treats the spatial properties of the scattering process as
described by the wave function. After an introduction, in Section XVII.2 the
differential equation for the radial wave function (the Schrodinger equation)
is derived. Section XVII.3 presents the solutions of the free radial wave
equation, and lists some of their properties. The properties of the exact
radial wave functions, in particular their asymptotic forms and their con-
nection to the phase shifts and the S-matrix, are discussed in Section XVII.4.
In Section XVII.5 the connection between bound states and poles of the
S-matrix on the positive imaginary axis is established. Some material about
functions of a complex variable, which is needed in this chapter and in
Chapter XVIII, is reviewed in a mathematical appendix.
XVII.1 Introduction
In the preceding sections we discussed those properties of the scattering

amplitude and the cross section that were consequences of very general
assumptions like unitarity and spherical symmetry. We derived relations
between the phase shifts, the inelasticity coefficients, the partial-wave ampli-
tudes, and the lth partial S-matrix elements. The cross section was ex-
pressed in terms of these quantities. Nowhere in our discussion did we
consider the spatial properties of the scattering process; in particular, we
did not have to make use of position-dependent quantities such as the
probability wave function.
425
426 XVII Free and Exact Radial Wave Functions
The probability wave function, however, serves to provide a pictorial

representation of the scattering process. In fact, it is in scattering theory
that the wave function has a direct relation to observation, since it represents
the probability for position measurements that can actually be performed.
(For stationary states such measurements are not performed in practice,
as has been discussed in the first half of the book; the wave function then is
merely an auxiliary mathematical quantity.)
In the present chapter we introduce the wave function. As in the previous
chapters, we consider two different points of view: Firstly, we assume time
development of the system as given by a Hamiltonian H = K + V. This will
provide a "deeper" foundation for the relations between the more directly
accessible quantities. We then see, however, that these relations have a more
general validity.
In addition to providing us with the wave function, this chapter will also
establish the connection between our more general coordinate-free formula-
tion of scattering theory and the usual formulation in terms of solutions of
the Schrodinger wave equation. This latter formulation is directly obtained
if the internal quantum numbers I] are ignored.
Though the Schrodinger differential equation constitutes a powerful tool
in many cases (described by an interaction potentia!), it will turn out that
the integral equation for the wave function (which is, in fact, just the
Lippmann-Schwinger equation in coordinate form) is of more immediate use.
XVII.2 The Radial Wave Equation
We start with the Lippmann-Schwinger equation in the angular momentum

basis
It is assumed that the position operator for projectile relative to target, Q,

commutes with the internal observables I]0P of the system:
By taking the scalar product of the Lippmann-Schwinger equation with the

generalized position eigenvector
we then get
<r e= <r e
where <r 8 ¢ rJ' IE 113 IJ) and <r 8 ¢ IJ' IE 113 IJ +) are called the jree and exact
wave junctions, respectively.
After scattering, the projectile's position x and momentum p have the same
direction (see Figure XIV.3.l):
x/lxl = p/lpl· (2.2)

Thus (8, ¢) are the angular coordinates of x as well as of p. On account of
spherical symmetry one can write, using the same arguments that led to
Equation (VII.3.l1),
<r8¢1J'IE1l31J) = <8¢11l3)<rlJ'IEIJ)1
= 1';1,(8, ¢)b~~.<rIE)7, (2.3a)
<r8¢1J'IE1l 3 1J+) = <8¢1 113)<rlJ'IEIJ+)1
= 1';/,(0, ¢)<r IJ'IE IJ+)I' (2.3b)
The reduced matrix elements Ffi<r IE)7 and Ffi<r IJ' IE IJ +)1 are called
the free and exact radial wave junctions, respectively.!
Using the Wigner-Eckart theorem for the scalar operator V, one obtains
<r8¢1J'IVIE1l 3 1J+) = _~ <r8¢11'IEii 3 f;+)<Eii 3 fj+IVIE1l 3 11+)
Ell,ij
= _I:
Ell,ij
<r8¢11'IEii 3 fj+)b 1f b"i,<Efj+IWIIEI1+)1
= ~ Yll/8,¢)<rlJ'IEij+)I<EIj+IWIIEIJ+)t
Eij
where <E fj+ IWIIE 11+)1 is the reduced matrix element of V. Thus
(2.3c')
where we have defined the reduced transition matrix element <r lJ'll VilE IJ+)I
in terms of the reduced matrix elements of V by
<r l1'IWIIE 1]+)1 = L

Eij
<r 11'1 E fj+ )I<E fj+ IWIIE 1]+ )1'
The reduced transition matrix element is a function of I, r, E, 11', and IJ only.

For spherically symmetric problems, V is an operator function of the radius
operator Q == (Q2)!/2. In general it also depends upon the internal ob-
servables, i.e., upon the operators that change the internal quantum num-
bers IJ.
We use the fact that [Q, V] = 0 to write
where <r I r) is the b-function with respect to the continuous summation
1 (r 1]'1 E 1]+), cannot, in general, be written as <rl E+ )"()"." because [H, 1]0P] = 0 might not be
true. Thus in general 1E 1] +) is not necessarily an eigenvector of 1]0P. Only if [H, 11°P] = [V, 1]0P] =
o is this true. Cf. the remark following Equation (XVI. 1.3).
over rand <'1'111 V(r)111 ii> is a doubly reduced matrix element depending only
on r, ii, and '1'. 2 Therefore
<r IJ'IIVIIE '1+ >1 = I <r 1J'llVllr ii> <r ii IE '1+ >1
rij
= Lrij <rlr><IJ'IIIV(r)lllii><riiIEIJ+>I'
Performing the summation over r and inserting the result,
<r IJ'IIVIIE '1+>1 = L <1J'IIIV(r)lllii><r iiIEIJ+>I'
ij
into (2.3c'), one obtains
<r e ¢ '1'1 VIE 11 3 '1+> = ~l/e, ¢) I <'1'111 V(r)III!?> <r iilE '1+>1'
ij
(2.3c)
Using the same arguments that led to (2.3c'), one can also show that
<re¢IJ'1 E _ ~ + iO VIE1l 3 1J+> = ~13(e, ¢)<rlJ'l E _ ~ + iO VIEIJ+>I'

(2.3d)
We shall now use the results (2.3) to obtain the Schrodinger differential
equation for the radial wave function. In Section XVII.3 we shall consider
the differential equation for the free radial wave function. Then, from the
Lippmann-Schwinger equation (2.1) for the wave function, we shall obtain,
in Section XVII A, the integral equation for the radial wave function. The
advantage of the integral equation over the differential equation is that the
former incorporates the boundary conditions; in this particular case the
boundary conditions appropriate to the Lippmann-Schwinger equation for
IE (j+ >are those for a free incident state, as discussed in Section XV.2.
We assume that the incident beam has projectiles of one kind, whose
mass m is not changed by the scattering process. The differential equation
is then obtained by taking the transition matrix element of the operator
H = K + V = p2 j2m + Kint(lJoP) + V, where P is the projectile momentum
operator:
E<re¢IJ'IE1l 3 1J+> = <re¢IJ'IHIE1l 3 1J+>

= <re¢IJ'I(p 2 j2m + K int + V)IE1l 3 1J+>
2m (~~
= _1 r2 dr (r2~)
dr + 1(1 +
r2 1) + 2mE int ) <re¢IJ'IE1l 3 1J+>
+ <re¢IJ'IVIE1l 3 1J+> (204)
2 This statement can be viewed as the Wigner-Eckart theorem for the group generated by Q
applied to the operator V, which is a scalar operator with respect to this group.
[The second equality follows from Equation (VII.2.6) and (VII.3.4) and is
the analogue of (VII.3.7).]
Using (2.3a, b, c) in (2.4), we obtain a system of coupled differential equa-
tions:
t 2
1 d ( d)
[{ r2 dr r dr -
1(1 + 1)
r2 + pij(E) 2} (j~,~
-2m<1J'IIIV(r)IIIQ>]<r11IE1J+>1 = 0, (2.5)
where
(2.6)
In order to solve this system of equations one has to know the reduced
matrix elements of the potential <1J'111 V(r)III1J>. These depend, of course,
upon the particular process under investigation. For instance, in the many-
channel problem of the elastic and inelastic scattering of electrons by atoms
these reduced matrix elements can be determined from the Coulomb inter-
action and the atomic wave function. 3
We shall not discuss the many-channel problem here any further 4 but
restrict ourselves to the case
(2.7)
Then the problem reduces to a one-channel problem for each value of 1J
(i.e., for each set of values of the internal quantum numbers). To perform
the reduction, we use the above arguments to write
<1J'111 V(r)IIIQ> = (j~'ij Vij(r). (2.8)
Since we now have [H, 1J oP ] = 0, we can also write

(2.9)
Under these conditions (2.5) goes over into a set of uncoupled equations,
one for each channel1J:
(2.10)
In the following sections we shall discuss the properties of the radial wave
functions <rIE+>7. Though y"(r) may vary with 1J, and p; may differ from
2mE = p2 by different constants 2mE~" the label 1J is inessential, as it does
not change in the scattering process. We shall therefore suppress it in most
of what follows, reinstating it at the end to emphasize the fact that further
quantum numbers may be present.
3 For the scattering of electrons by hydrogen atoms this is discussed in Massey et al. (1969),
Vol. I, Section 7.1. A more general case is treated in Smith (1971), Section 2.1.
4 We shall return to the many-channel problem in Chapter XX.
XVII.3 The Free Radial Wave Function

The equation for the free radial wave function <r I E), is obtained from the
eigenvalue equation of K:
E<r8cprt'IE1l 3 1J) = <r8cplJ'IKIEI1 3 1J)
in the same way as (2.5) is obtained from (2.4). It differs from (2.5) only in
that the term in V is missing and that since [K, lJ oP ] = 0,
(3.1)
We thus have
V2:r(r2fr) _l(l :2 1) + p2}<r IE)7 = 0, (3.2)
where p = p~(E) == J2m(E - E~nt). We shall write <rIF)? = <rip)? Equa-

tion (3.2) is a well-known equation, which has as its linearly independent
solutions the spherical Bessel and Neumann functions. The solution that
is regular at r = 0, to which <r I P), is proportional, is the (proper) spherical
Besselfunction (spherical Bessel function of the first kind, or Riccati~Bessel
function):5
n )1/2
j,(pr) =j,(z) = ( 2z J , + 1/2(Z)
_
-(-z ~~ - z)
)1(1 d)' (sin
z dz z
_ 1 00 (_z2j2)2
L
- z n=on.'(21 + 2n+ 1) II
.. '
pr = z, (3.3)
°
and the solution irregular at r = is the spherical N eumannfunction (spherical
Bessel function of the second kind or Riccati~ Neumann function):
-1-1 ~ (z2j2)n(21- 2n - 1)!!

= -z ~ , . (3.4)
n=O n.
Here J , + 1/2(Z) and J -1-l/2(Z) are ordinary Bessel functions of the first kind.
5 The double factorial is defined by
n!! == {n(n ~ 2)·· ·(5)(3)(1) if n is odd,

n(n ~ 2) ... (4)(2) if n is even.
For further properties and of the spherical Bessel functions and their proofs see Abramowitz
et a/. (1965). Chapter 10. and Watson (1958).
XVII.3 The Free Radial Wave Function 431
The spherical Hankelfunctions, defined by

hz(z) = iz(z) + inl(Z), (3.Sa)
ht(z) = iz(z) - inz(z), (3.Sb)
are then also independent solutions of (3.2). We note for later use that all the
above relations also hold for complex values of z.
The asymptotic behavior of iz(z), nl(z), and htCz) for real z -> 00 is
iz(z) '"
z-oo z
~ sin (z - In)
2
, (3.6a)
nlz) '" - 1
-cos ( z - -In) , (3.6b)
z- 00 z 2
( - I.)1+ 1
he)
lZ'"
iz
e. (3.6c)
z- 00 Z
For future reference we note the bounds for complex values of z:

( 1_)1+1
I iz(z)1 s const ~ ellmzl, (3.6d)
IhtCz)1 s
1
const ( 1 + Izl
)-1 e- 1mz
• (3.6e)
Near the origin we have

Zl
(3.7a)
iz(z) z~0(21 + I)!!'
nl(z) -> [-(21- I)! !]Z-I-1. (3.7b)
The normalization of the j,'s is
Jrooo r2 dr iz(pr)jl(p'r) = 2p2

n
b(p - pi), (3.8a)
roo n
Jo p2 dp iz(pr)iz(pr') = 2r2 b(r - r'). (3.8b)
The spherical Neumann function cannot be delta-function normalized

in this way. All spherical functions iz(z), n,(z), h1(z), and ht(z) fulfill the
functional equations
d 1
dziz(z) = 21 + 1 [/iz-1(z) - (l + l)iz+ 1(Z)] (3.9a)
and the following equality for the indefinite integral:
f r2 dr fz(pr)f[(p'r) = p 2 r2
- p
12 [p1fz(pr)fi-1(p'r) - piz-1(pr)ft(p'r)],
(3.9b)
of which (3.8) is a consequence.
To find the proportionality factor between <r IP)l and jz(pr), we calculate
the normalization integral of the free radial wave function <rlp)l' In spherical
coordinates the J-function normalization of the generalized position eigen-
vectors reads
<xix') = J3(X - x') = ~() J(r - r')J«(} - (}')J(¢ - ¢'); (3.10)

r sm
Inserting the sum over the complete system of basis vectors gives:
<x Ix') = L fp(E) dE <r () ¢ IE 11 3 ) <E 1131 r' ()' ¢')

113
= L fp(E) dE Yzl/(}, ¢)<rIE)1 ¥t3«(}', ¢')<r'IE){

113
= ~()
sm
J«(} - (}')J(¢ - ¢') fp(E) dE <rIE)I<r'IE)j,
where we have used

I
L Yzl/(}' ¢)Yl~((}"
~
¢') = --;--(}b«(} - (}')J(¢ - ¢').
~n
(VII.3.19)
Comparing this with (3.10) shows that the normalization of the radial wave
function is
fp(E) dE <rIE)z(r'IE){ = rl2J(r - r'), (3.11 )
or, in terms of p,
f p2 dp <rlp)l<r'lp){ = r12 J(r - r'). (3.11')
As <r IP)l is a solution of (3.2) regular at r = 0, and therefore proportional

to jz(pr), one obtains by comparison of (3.11') with (3.8b) that
<rlp)l = yl2fii.jz(pr). (3.12)
XVIIA The Exact Radial Wave Function
The precise properties of the exact radial wave function <r I E +)1 depend
upon the interaction Hamiltonian V, through the interaction potentials
V~(r) = <x 1] IV Ix 1]). However, provided V fulfills some very general con-
ditions, one can make some general statements about <r IE+)l without
knowing the detailed form of V. The content of these conditions is roughly
as follows: 6
1.
°
The interaction is not effective for large projectile-target separation,
i.e., VCr) --> as r --> 00. Which precise assumption one has to make
6 See, e.g., Taylor (1972).

XVlI.4 The Exact Radial Wave Function 433
for the rate of falling off of V depends on what one wants to achieve:
Usually one assumes that VCr) falls off at least as fast as 1/r 3 • For the
derivation of some properties one has to assume that VCr) decreases
faster than any power of 1/r, and often one assumes a finite range for V.
2. The interaction does not change too abruptly as the projectile ap-
proaches the target: VCr) is a continuous function of r except at a
finite number of finite discontinuities.
3. There exists a lowest energy value (since the binding energy must be
finite) or, in other words, the spectrum of H has a lower bound. In terms
of the potential this means that VCr) is not too singular at the origin
r = 0 (less singular than r- 3 / 2 if VCr) < 0).
Almost all physical potentials fulfill these conditions-potentials for
electrons scattering off an atom, potentials for atom-atom scattering, the
Yukawa potential, the square-well potential, etc. The Coulomb potential
does not fulfill condition 1 and therefore requires a separate mathematical
treatment. In practice, however, this is not of much consequence, because
electromagnetic screening is always present: One never really has the exact
Coulomb potential 1/r, but a screened potential Ij;(r)/r, with Ij;(r) = 0 for
large r.7 For the following discussion we assume that the above three con-
ditions are fulfilled in whatever precise form we need.
If
VCr) > 0 for all r, (4.1)
then the potential is said to be repulsive, while if
VCr) < 0 for all r, (4.2)
then it is said to be attractive. 8 If V(r) < 0 in some interval r 1 < r < r 2' then
VCr) is said to be attractive in that interval. If VCr) is attractive in some
interval, there may be a certain number of solutions of the differential equa-
tion (2.10) for a certain number of discrete, negative values of p~/2m = En.
These solutions correspond to bound states of the projectile-target system
with binding energies IEnl, as we shall discuss in the next section. They must
fall off sufficiently rapidly as r --+ 00, in addition to being regular at r = O.
The exact radial wave function <r IE+)I = <r IP+)1 for p2 ~ 0 is the
solution of the differential equation (2.10) with certain boundary conditions:
In addition to being regular (i.e., <r Ip+) < 00 at r = 0), one would want, as
r --+ 00, <rlp+)1 to represent the Ith spherical component of an incident
plane wave and an outgoing scattered wave, and that for V --+ 0, <rip + ) 1 --+
<rlp)1 = fificHpr), the acceptable solution of the free wave equation.
Instead of solving (2.10) under these boundary conditions (see Problem
1), we shall obtain the exact radial wave function <r IP+)1 from the
Lippmann-Schwinger equation (2.1), which has the boundary conditions
7 Note also that in general any assumed V(r) is a mathematical idealization of the actual
situation.
8 The forces between projectile and target are then also called repulsive and attractive,
respectively.
already built in. Inserting (2.3a), (2.3b), (2.3d), (2.8), and (2.9) into (2.1), one
obtains the integral equation for the radial wave function <rIE+)I:
<rIE+)l
= <rIE)l + I peE') dE' r'2 dr' <rIE')l E _ ~, + iO <rllE')iV(r')<r'IE+)l'

(4.3)
Now define the (lth-partial-wave) Green's function by
GP(r' r') = _1 fa) (E') dE' <rIE')l<r'IE')i (4.4)

l' 2m 0 p E - E' + iO
_ fa) 12 d <r I p')l<r'l p')i
(4.5)
I
- o p P P2 - P12 + I'0
_ ~ fOO 12 d I Hp' r)Hp' r')

- n 0
p p p 2 -p 12 + I'0 (4.6)
- ~ f+ a) 12 d 1t(plr)Hplr') I
- n -a) p p p 2 -p 12 + I'0' (4.7)
where we have used the fact that the integrand is even. The integral equation
(4.3) then becomes
<rlp+)l = <rlp)l + 2m Ioa) r'2 dr' Gf(r; rl)V(rl)<r'lp+)l
= j2!;:,1t(pr) + 2m La) r'2 dr' Gf(r; rl)V(rl)<r'lp+)l' (4.8)
Note that if the potential has a finite range, then the integral is over a finite
interval.
The Green's function is evaluated using Equation (3.5) to write
Gf(r; r') = ~ J+a) dq q2 1t(qr)h l(qr? + ~ f+a) dkk 2 1t(kr)h i (kr? (4.9)
2n _ a) p2 - q2 + If 2n _ a) p2 - k 2 + If
Suppose r' > r. Then because of the asymptotic behavior (3.6c) of hI> and
(3.6d), the product 1t(qr)hl(qr') decreases exponentially in the upper half
complex q-plane, and one may extend the integration path of the first integral
to include a large semicircle in the upper half plane (cf. Figure 4.1). From
Equations (3.3), (3.4), and (3.5) it is seen that
1t(kr) = (-l)Jl( -kr) and hi(kr') = (-l)lhtC -kr')
and hence with the substitution q = - k, the second integral of (4.9) is the
same as the first integral. We may now rewrite (4.9) as
G1P( r,. r ') -- l'1m --2

R~a) 2n
f [(R)
d
q
(p
q2jtCqr)hl(qr') . ,
+ q)(q - p - If)
where r(R) is the closed contour indicated by Figure 4.1. The spherical
XVII.4 The Exact Radial Wave Function 435
1m q
--~=======t==~====~-- Req
-R +R
Figure 4.1 The integration path rcR) used to evaluate the first integral in Equation
(4.9). As R --> 00 , the contribution from the semicircle vanishes, leaving only the integral
along the Re q axis.
Bessel functions are entire functions, and the function q2jz(qr)hzCqr')j(p + q)

is analytic inside f(R) , so by the Cauchy residue theorem 9 we obtain
2
Gr(r; r') = - 2i -p- Hpr )hlpr')
p+p
= -ipHpr)hz(pr') (r' > r). (4.l0a)
A similar procedure, but with rand r' interchanged, gives the Green's function
for r' < r:
Gr(r; r') = - ipj/(pr')hlpr) (r' < r). (4. lOb)
Equations (4.10) are often combined as
Gr(r; r') = - ipj/(pr < )hz(pr » (4.11)
where
r < = miner, r') (4.l2a)
and
r> = max(r, r') . (4.l2b)
We already know the asymptotic behavior of the free radial wave func-
tion <rlp ) / = fiFrHpr) for r --> Cf) [cf. (3.6a)]. Let us now examine the
asymptotic behavior of the exact radial wave function <rlp+ ) /. Using (3.6a),
(3.6c), and (4.l0b) in (4.8), we obtain
A - <rip + )/ ~ -1 sm
2 r~ oo pr
. (pr - -
2
In)
+2m y2(n f oo r'2dr' [ -ipj/(pr') (-iy+
0 pr 1 . ]
e1pr V(r')<r'lp+ ) /
= -1
2ipr
{e- i (pr - /1[ !2)
- ei (pr-/rr/ 2 { 1 _ (2ip)( J 2nm) 50 00

r'2drIHprl)V(rl)<r/IP +)/]}
(4.13)
9 For a review of important results in the theory of functions of a complex variable, see
Appendix XVII.A.
The first term represents an incoming partial wave, whereas the second
term represents an outgoing one. (We shall discuss this further at the end of
this section.) The outgoing wave is modified with respect to the incoming
one by the factor in square brackets. This factor may be related to the elastic
scattering phase shift defined by Equation (XVI.2.7) for every value of the
internal quantum number Yl. Note that having chosen the quantum numbers
Yl such that (2.7) holds, <YlbIISl(E)IIYlA) vanishes for Ylb i= I]A' Thus we have
here only elastic scattering for every value of 1].
Recall, now, that the lth partial-wave amplitude defined by (XVI.l.35),
(XVI.1.18), and (XVI.1.23) is
1/(p) = -nm<E lI31 TIE lI3) = -nm<E lI31 VIE lit), (4.14)
where E = E(p) = p2 /2m + Eint, the label I] having been suppressed. In
terms of the radial wave function this is written
1/(p) = -nm fr2 sin 0 dr dO d4> <E lI31r 04» <r 0 4>JVIE lit)
= -nm[fSin 0 dO d4> YtiJ(O, 4»Yll /O, 4»J
x [fr2 dr <rIE)tV(r)<rIE+)I]
1/(p) = -nm f' r 2 dr tNpr)V(r)<rIP+)I' (4.15)
where Equations (2.3a), (2.3c), (2.8), (2.9), (VII.3.16), and (3.12) have been
used. Using (XVL1.47) and (XVI.2.7), it follows from (4.15) that
Slp) = e2ibl = 1 + 2ip1/(p)
(p)
= 1 - (2ip)(fom) {"" r2 dr Npr)V(r)<rlp+)I ' (4.16)
Although Equations (4.15) and (4.16) are important in themselves, ex-

pressing the lth partial-wave amplitude and the phase shift in terms of the
potential and the exact radial wave function, they are of little practical use.
This is because the differential equation (2.10) and the integral equation
(4.8) are not easily solved for <rlp+)l' nor is <rlp+)l as directly related to
the experimental data as are <5lp) and 1/(p).
( r lp >o -sin pr
,.\f
••
ror r > a
Figure 4.2 The I = 0 free and exact radial wave functions for an attractive square well.
XVIIA The Exact Radial Wave Function 437
We make use of (4.15) and (4.16) to rewrite (4.13) in various useful forms:
1 (In) ei(pr-17,/2)
Jn72<rlp+)1 '" - sin pr - -2 + p~(p) (4.17a)
r-+ 00 pr pr
'" _~_ {ei(pr-I1t/2)e2ibr(p) _ e - i(pr-I1t/2)} (4.17b)

r-+ 00 2zpr
(4.17c)
(4.17d)
Comparison of (4. 17c) with the asymptotic form of the free wave function
obtained from (4.12) and (3.6a),
A "2 <rlp)l = Hpr) '" -1 sm

r-+ 00 pr
In) '
. ( pr - -2 (4.18)
explains why Ci/p) is called the" phase shift": The interaction shifts the phase
of the asymptotic exact radial wave function <r Ip +)1 by Ci/(p) as compared
with the asymptotic free radial wave function <rlp)/. This property also
serves as an alternative definition of Ci/ in place of (XVI.2.7). The phase shift
for the I = 0 wave in a square-well potential is shown in Figure 4.2.
For an attractive potential (like the square well in Figure 4.2) the wave
function is "pulled" into the interaction region and the phase shift is positive.
The exact wave function emerges from the interaction region with its phase
in advance of the free wave function. For a repulsive potential the wave
function is "pushed" out of the interaction region and the phase shift is
negative.
That this connection between the sign of the phase shift and the sign of the
potential is generally fulfilled is best established in the small-phase-shift
approximation:
eibr(p) sin CitCp) ~ CitCp)·
From (4.15), using (XVI.2.8), one then obtains
Ci1(P) ~ -nmp Loo

r 2 dr.)VnHpr)V(r)<rlp+)I,
which in the Born approximation is given by
CitCp) ~ -2mp 1 00
r2 dr UtCpr)}2 VCr).
To fully justify the above statement that (4.13) and (4.17) represent an
incoming and an outgoing partial wave, we have to consider the time-
dependent wave function of a state (V(t) that develops in time from a
prepared in-state.
The exact radial wave function <r IP+)1 is the pth component of the radial
wave function of an angular-momentum eigenstate (M 3 (t) that develops in
time according to the equation
cfJlt(t) = f
21n dE $(E)I E llt )e- iEt, (4.19)
where $(E) == $(p), E = p2/2m, is the energy (momentum) distribution of

this state in the remote past. iO The wave function of this state is
(4.20)
where he radial wave function <rlcfJ+(t)1 of the state cfJ+(t) is given by
<rlcfJ+(t)1 = 2~ {O dE $(E)<rIE+)le- iEt • (4.21)
Multiplying (4.17d) by j2!Tc(1/2n)$(E)e- iEt and integrating over E then

gives
(4.22)
Thus the radial wave function of the angular-momentum eigenstate

cfJlt(t) consists asymptotically of an incoming wave with the original momen-
tum distribution and an outgoing wave in which the original momentum
distribution has been modified by SI(P) due to the interaction. Equations
(4.13) and (4.17) merely state this result in the time-independent form for
the pth component of the radial wave function, i.e., for an (unphysical)
exact momentum eigenstate.
The asymptotic forms of the radial wave function (4.22) and (4.17) have
been derived from the Lippmann-Schwinger equation under the assumption
that the interaction is described by an interaction Hamiltonian V. Then
SI(P) is given in terms of VCr) by (4.16). Even if the time-development axiom
is not assumed, there exists a unitary operator S that transforms states before
the interaction into states after the interaction. If this interaction is of finite
range, then outside the interaction region the wave function should still
be a superposition of incoming and outgoing waves. Thus the radial ~ave
function for large distances r should still be given by (4.22), where SI(P), the
lth S-matrix element, is now the fundamental quantity. Consequently, the
10 That this (p+ (t). connected with the basis vectors IEll;), is a state that develops from a
prepared in-state has been shown in Section XV.2 and is here of no further consequence.
XVII.S Poles and Bound States 439
pth component of the radial wave function, <r Ip+ )/' should have the asymp-
totic form (4.17).
Thus in the framework of a Hamiltonian time development, the asymptotic
forms (4.17) and (4.22) are derived, while in the framework of S-matrix
theory the asymptotic forms (4.17) and (4.22) are assumed and are justified
as a consequence of the superposition principle for the wave function out-
side the interaction region.
XVII.S Poles and Bound States··
In scattering processes, the variable p that we have been considering in the

preceding section is the magnitude of the incident momentum, and therefore
a real positive quantity. Accordingly, the variable E = p2/2m representing
the kinetic energy (or E - E~t = p2/2m if the internal energy has to be taken
into account) is a real positive quantity. We know, however, that other
values of E also have physical significance. For example, the solutions of
(2.10) for negative values of p2 = 2mE are the bound-state radial wave func-
tions, and the discrete negative energy values En, for which (2.10) has normal-
izable solutions <r IEn)/, are the energy levels of the projectile-target bound
system (as has been discussed in Section VII.3).
One should then also be able to extend the integral equation (4.8) [with
Gf given by (4.11)], which follows from (2.10), to negative-energy solutions.
Such solutions can be obtained by the following line of reasoning: Replace
p by the complex variable z. For 1m z ~ 0, the arguments that lead to the
Green's function (4.11) are still valid. (For 1m z < 0, similar arguments,
wherein one closes the contour in the lower half of the complex z-plane,
hold, and one gets analogous expressions with hi replaced by ht.)
The fact that (4.11) holds in this general case does not necessarily mean
that (4.8) can also be continued to 1m z > 0. Let us assume, however, that
the potential VCr) is such that it can be so continued, and let us assume that
H has a bound state with energy E = - rx 2 /2m. Then, at the point z = irx,
the radial wave function <r Iirx+)1 (being, as a wave function of a bound state,
an element of the space of infinitely differentiable, rapidly decreasing func-
tions that is a realization of the Schwartz space) must asymptotically
decrease faster than any power of r - 1, e.g., it must decrease exponentially.
Now according to (3.6d), jlirxr) diverges exponentially. The second term in
the right-hand side of (4.8) is bounded as r --> 00, because hlirxr) decreases
exponentially according t~.6e). Thus <r Iirx+)1 can decrease for r --> 00
only if the term <r Ip)1 = -j2InHpr), originating according to (4.13) or (4.17a)
from the incident state, drops out of Equation (4.8). When can this be the
11 For this section, as well as for some sections of Chapter XVIII, we shall require some
basic results from the theory of functions of a complex variable. These results are outlined in a
mathematical appendix to this chapter, Appendix XVII.A. The reader who is unfamiliar with
complex variable theory, or who merely wishes to refresh his memory, may consult the appendix
before reading this section.
case'? In order to check this we consider, instead of (4.8), the equation

obtained by multiplying Equation (4.8) by (z - irx):
(z - irx)<rlz+)l = (z - irx)<rlz)l
-2izm L oo
r'2 dr' Nzr <)hbr>)V(r')(z - irx)<r'lz+)I'
(5.1)
Equation (5.1) is identically fulfilled at z = irx unless <rlz+)l has a pole
at z = irx, i.e., unless in the neighborhood of irx it is of the form 12
+ 1 2
<rlz )1 = - - . <rlE = -rx /2m)1
z - lrx
+ (function of z analytic around z = irx). (5.2)
With the ansatz (5.2) for <rlz+)l one obtains from (5.1) at z = irx
<rlE = -rx 2/2m)1 = 2mrx L oo

r,2 dr' Nirxr <)h1(irxr»V(r')<r'IE = -rx 2/2m)/.
(5.3)
In (5.3) the troublesome term <rlp)l has disappeared and we have obtained
a homogeneous integral equation for the discrete number of normalizable
solutions of the Schr6dinger equation (2.10) with eigenvalues E = -rx 2 /2m.
Thus we see that a bound-state solution is obtained if we assume that <r Iz +)
is of the form (5.2), i.e., that it has a pole on the positive imaginary momentum
axis at the point z = irx = iJ2m IE I, where E is the bound-state energy eigen-
value for angular momentum I. The bound-state radial wave function
<r IE = - rx 2 /2m)/ is then the residue at this pole.
From (4.15) and (4.16) one then concludes that the lth partial-wave am-
plitude and the Ith S-matrix element have poles on the positive imaginary
momentum axis at the values P = iJ2m IE~ I, where E~ (n = 1, 2, ... ) are the
discrete eigenvalues of the Hamiltonian H. In a theory with Hamiltonian
time development, these discrete eigenvalues correspond to the energy levels
of the projectile-target bound states of angular momentum l.
Though the above arguments 13 show that a bound state of the Hamiltonian
with energy E~ corresponds to a pole of the Ith partial S-matrix element
S e(P) on the positive imaginary axis, the reverse need not hold: ll(p) and
Seep) may have singularities that are not connected with the bound states. 14
Nevertheless, the hypothesis that bound states of angular momentum 1
correspond to poles of SzCp) on the positive imaginary axis (and vice versa)
12 The above arguments will also hold for a pole of higher order, but one can prove that these
poles are simple (which we shall not, however, do here).
13 For potential scattering, the connection between simple poles and bound-state eigen-
values of H can be made more precise. Cf. Taylor (1972), Chapter 12.
14 These "redundant" poles may occur if the asymptotic expressions in the complex p-plane
differ from (4.17). For interactions that are cut off at a finite distance and also for potentials
that fall off at infinity faster than any exponential (e.g. e- W ), "redundant" poles are absent.
XVII.6 Survey of Some General Properties of Scattering Amplitudes 441
has become quite generally accepted. Especially in those cases where a

Hamiltonian time development is not assumed to exist (relativistic S-matrix
theory) and there is no operator whose spectral properties characterize the
bound states, one usually assumes a one-to-one correspondence between
bound states and poles on the positive imaginary axis.
XVII.6 Survey of Some General Properties of Scattering

Amplitudes and Phase Shifts
The precise properties of the scattering amplitudes and the phase shifts
depend upon the interaction. If the interaction is described by a potential,
one can calculate phase shifts and partial-wave scattering amplitudes as a
function of scattering energy, as has been done in some of the problems.
There are, however, general properties of these functions that do not depend
upon the particular form of the interaction, and we want to list them here.
The integral equation for the radial wave function can be iterated, thereby
yielding:
<rlp+)l = <rlp)l + 2m Jr'z dr' Gf(r,r')V(r')<r'lp)l
+ (2m)2 Jr'z dr' Gf(r, r')V(r') J r"2 dr" Gf(r', r")V(r")<r"lp)1
+ ...
Inserting this into (4.15) one obtains a series which contain powers of the
potential and the spherical Bessel functions. For weak potentials and for
high energies the higher-order terms in the potential and in the spherical
Bessel functions are small (due to the property (3.6a) for the spherical
Bessel functions). For this reason the Born approximation ofthe partial-wave
amplitude, obtained from (4.15) by restricting the above expansion to the
lowest -order term:
TjBorn)(p) = 2
-nm- foo r2 dr jz(pr)V(r)jz(pr), (6.1)
n 0
is believed to be a good approximation for high energies and weak potentials.

The asymptotic behavior of the lth partial S-matrix element is then
(6.2)
This follows immediately from (4.16) using the Born approximation (6.1)
and the asymptotic behavior (3.6a) of jz(pr). An intuitive argument for (6.2)
is that with increasing energy of the projectile the effect of a given interaction
will become less important.
The phase shift c5/p), therefore, tends to a multiple of n as P -> 00. As the
phase shift is defined by (XV1.2.7) only up to a multiple of n, one can remove
this modulo n ambiguity by defining
c5z(p)->O as p->oo. (6.3)
Requiring then that bl(p) be a continuous function of p (which is possible

because Slp) is continuous) makes blp) unique.
Clearly Slp) ---+ 1 also if the interaction goes to zero. Then with the con-
vention (6.3) it follows that also
bl(p) ---+ ° for interaction going to zero eVer) ---+ 0]. (6.4)
For a given potential and given energy
7; ---+ ° and SI ---+ 1 as I ---+ 00. (6.5)
Intuitively, this can be understood from (2.10) by regarding the term

l(l + 1)/2mr2 as a repulsive centrifugal potential. The larger I is, the more
repulsive is this centrifugal barrier and the less effective does the actual po-
tential VCr) become. Let R be the range of the interaction. Then for values I
such that l(l + 1)/2mR2 is much larger than the kinetic energy E = p2/2m
the projectile is unlikely to penetrate into the range of the interaction. Thus
for
the partial-wave amplitude 7;(p) will be negligible:
7;(p) ~ 0, Slp) ~ 1, for I ~ pRo (6.6)
It follows from (6.6) that for the phase shift
n integer, for 1 ~ pR. (6.7)
Equations (6.6) and (6.7) are important properties of which one makes use
when one expresses the differential cross section in terms of the partial-wave
amplitudes as in (XVI. 1.36). For a given value of E = p2/2m the infinite
sum in (XVI.1.36) can, because of (6.6), be approximated by a finite sum.
To obtain the phase-shift behavior at low energies, we turn to (4.15). For
small values of p, <r Ip+)1 and Hpr) have the same p-dependence, given by
(3.7a), since according to (4.11),
GP( ') . Plr l< (plr l>·(21 1) " -1- 1 -1- 1 )

Ir,r ---+-IP(21+1)!! (21+1)!!-1 - .. p r>
°
1 -1-1
GP( ')
I r, r ---+-
_r_<r_>__ (6.8)
for p ---+
21 + 1
and <rlp+)1 andjl(pr) are connected by (4.8). Consequently, it follows from
(4.15), for potentials that vanish for r greater than some R, that
7;(P)---+ _nm(f R r 2 dr grIV(r)r l 1 2 <r IP +)I)p21

Jo ~~ [(21+ I)!!] Hpr)
7;(p) ---+ - al p21 for p ---+ 0, (6.9)
XVII.6 Survey of Some General Properties of Scattering Amplitudes 443
where al is a constant given by the integral in parentheses, which does not

depend upon p, because <rlp+ >dNpr) ---+ (function of r) for p ---+ 0. 15
The constants a1 are called the scattering lengths. Only the s-wave scatter-
ing length ao has the dimension of a length. (6.9) shows that for p = 0 all
partial-wave amplitudes except the s-wave vanish. For the s-wave
To(p) ---+ - ao as p ---+ 0, (6.10)
and consequently by (XVI.1.36) and (XVI.1.38),
da 2
dO. ---+ ao, a ---+ 4na5, as p ---+ 0. (6.11)
Using (XVI.2.8) it follows further that

bl(p) - nn ---+ -aIP21+ 1 as p ---+ 0, (6.12)
and consequently also

tan blp) ---+ -aIP2l+ 1 as p ---+ O. (6.13)
Equation (6.13) shows that
°
or their inverses are the appropriate functions to consider near p = (as °
°
long as the scattering lengths al #- or al #- 00). The power series for this
quantity near p = is:
p21+ 1 cot bl(p) = __1 + -.!r p2 + O(p4). (6.14)

al 2
That only even powers appear in the power-series expansion of p21+ 1 cot blp)
follows from bl( - p) = -bl(P), which is established in Equation (XVIII.S.7)
below.
For I = 0, (6.14) is a very useful approximation called the effective-range
approximation:
ro 2
pcotbo(p) = - -1 + -p. (6.15)
ao 2
The constant ro is called the effective range of the potential; one can show
that ro is roughly proportional to the range of the potential. Equation (6.15)
is a good approximation for energies small compared to the potential energy
and has been one of the most important parameterizations of low-energy
scattering data, in particular for neutron-proton scattering.
15 If the potential does not have a finite range, some modifications are necessary, but one
can show that for VCr) decreasing faster than any power of Ifr the result (6.9) still holds.
XVII.A Mathematical Appendix on Analytic Functions
In this appendix we review some basic results in the theory of analytic func-
tions of a complex variable. For proofs of these results, and for more on the
theory of analytic functions, the reader is referred to, e.g., Smirnov (1964),
Vol. II, Part 2.
a. Definition of Analyticity
The function fez) = u + iv of the complex variable z = x + iy is said to be

analytic in some region R of the z-plane when
1. dfldz is continuous and independent of the direction of dz in the
region R. (The following conditions are equivalent.)
2. au/ax = av/ay, au/ay = - av/ax (Cauchy-Riemann conditions). These
derivatives are also continuous in R.
°
3. ~c fez) dz = for any closed contour C lying within R.
If fez) is analytic in R, then so is dnfldz n for all n = 0, 1,2, ....
b. Cauchy Integral Formula
If fez) is analytic within some simply connected region R, then
1 fez) dz = 2nif(a), (A.la)

J z- a
1 fez) dz = 2ni f(n)(a) == 2ni c!'.f(z~ I (A.1b)
J (z - at+ 1 n! n! dz n Z= a
for contours entirely within R and enclosing the point a. If a is excluded, the
integrals vanish. (Integration is assumed to be counterclockwise.) Thus the
values of fez) within R are completely determined by the values on the
boundary.
c. Taylor Series
If fez) is analytic within and on a circle C centered at z = Zo, then one can
expand fez) about Zo :
ao
fez) = L an(z - zot, (A.2a)
n=O
where
_ 1 f(n)( ) (A.2b)
an - , Zo'
n.
XVII.A Mathematical Appendix 445
This Taylor series is uniformly convergent as long as the radius of C is less

than the "radius of convergence," which is the distance from Zo to the nearest
singularity of fez), i.e., the nearest point where f is nonanalytic. Conversely,
any power series represents an analytic function within its radius of con-
vergence.
d. Laurent Series
If fez) is analytic within and on the annular region between two concentric
circles C 1 and C 2 centered at z = zo, then for any point z within the annular
region one has the uniformly convergent power series
00
fez) = L an(z - zo)", (A.3a)

n= - 00
where
(A.3b)
C being any contour within the annulus.
e. Singularities
Points of nonanalyticity of f are called singularities of f.

1. A single-valued function f can have isolated singularities of the fol-
lowing kinds:
(a) A singularity at z = a is called a pole of order n if as z -+ a
fez) -+ g(z) (n = 1,2, ... ), (A.4)
(z - a)"
where g(z) is analytic at z = a, and g(a) of. O. A pole of order 1
is called a simple pole.
(b) If f(z)(z - a)n diverges as z -+ a for all finite n, then a is called
an essential singularity of f.
In the Laurent series for f about a pole of order n, - n is the lowest
power one encounters. The Laurent series about an essential singular-
ity contains infinitely many terms of negative power.
2. A multivalued function f (e.g., f = Z1 / 2) has branch points, which
always occur in pairs. At a branch point z = a, the functionf(a + te itP ),
where f is such that a + te itP does not include another branch point of
ffor all cp, is not periodic in cp with period 2n.
A function that is analytic in a region of the complex plane, except for a
finite number of poles in that region, is said to be meromorphic in that region.
If the function is meromorphic in the entire z-plane, except at the point 00,
it is called a meromorphic function of z. If the function is analytic in the entire

z-plane, except at 00, it is called an entire function.
f. Cauchy Residue Theorem
The coefficient a_I in the Laurent series is called the residue of fez) at
z = Zo. If fez) is analytic within and on a closed contour C, except for a
finite number of poles and essential singularities, then
£ fez) dz = 2ni I (residues in C) (A.S)
g. Analytic Continuation
Let two functions be analytical in a region R, and have the same values (a)
in some subregion of R, or (b) on a line segment in R, or (c) at a denumerably
infinite number of points having a limit point within R. Then they are
identical throughout R. However, if one knows only that they differ at most
by an amount f in (a), (b), or (c), then, no matter how small If I is, the two
functions may differ vastly within the rest of R.
The facts stated above illustrate the justification and the limitations of
analytic continuation: If f1(Z) is defined on a line segment r, and fiz) is
analytic in a region R containing r, and iffl(z) = f2(Z) on r, thenf2(z) is the
unique analytic continuation of f1(Z) onto R.
In the same spirit one can also do the following: Let fl (z) be given within
a circle R 1 , e.g., by a power series about the center ZI. Letf2(z) be the power-
series expansion offl(z) about a point Z2 near the periphery of R 1 , the circle
of convergence being R 2 , which extends beyond R I . Then the values of
fiz) in R2 are uniquely determined by the values of fl(Z) in R 1 , and f2(Z) is
the analytic continuation of fl (z) into R 2. One can repeat this process
several times, obtaining a unique analytic continuation F(z) of fl(z) into
the region Rl u R2 U R3 u ... (see Figure A.1). A function cannot be
analytically continued into a region containing a singularity of that function.
After repeating the process of analytic continuation several times, the
nth circle of convergence may overlap the first one. Then the values of fn(z)
etc.
Figure A.I Extension ofa function defined in Rl into the region R2 u R3 U R4 U···
by analytic continuation.
branch
point ------:>"f-,,,.c;--,~~-"i.
Figure A.2 Analytic continuation of a multivalued function around a branch point.
in the overlap region mayor may not coincide with those of II (z). In the
latter case the function is multivalued, and the region of encirclement (see
Figure A.2) contains a simple branch point of the function.
We have seen above how the values of an analytic function in a large
region of the complex plane are determined by its values in an arbitrary
small region of analyticity. However, the analytic continuation does not
depend on these initial values in a continuous manner. Therefore, it is im-
possible to express F(z) in R explicitly in terms of the values in a certain
small region. Though F(z) is uniquely prescribed by its values in the sub-
region, the problem of constructing F(z) in R from its values in the small
subregion is not a correctly formulated one.
h. Schwarz Reflection Principle

If I(z) is analytic in a region R that includes a segment of the real axis and
if fez) is real on this segment, then fez) can be continued onto the region
R* = {z* Iz E R} and satisfies fez) = [f(z*)]* for all z E R u R*.
i. Multivalued Functions
The function w(z) can be considered as a map of the z-plane onto the w-plane.
Consider the example w(z) = Z2, Figure A.3. It is clear from the figure that
p'
p q' T'
(a) (b)
Figure A.3 The function w(z) = Z2 (a) z-plane; (b) w-plane.
the upper half of the z-plane, including the positive real axis and excluding
the negative real axis, is mapped onto the entire w-plane. So also is the lower
half of the z-plane, including the negative real axis and excluding the positive
real axis. Thus one complete circuit of the z-plane (e.g., efgpqr) involves two
complete circuits of the w-plane (e.g., e'f'g'p'q'r') provided the former circuit
encloses the origin.
The inverse of a function such as w(z) is a multivalued function. To con-
tinue with the above example, consider the inverse "function"
fez) = ZI/2.
Writing z and f in polar form,

z = pe i</>,
we have
e= </J/2.
In making one complete circuit around the origin of the z-plane, 0 S </J < 2n,
one has
fl = pl/2e i</>/2,
and the part of the f-plane so covered is the upper half including the + Re f
axis and excluding the - Re faxis. Going around the z-plane once more
(2n S </J < 4n), the point with polar angle </J now has polar angle </J + 2n;
consequently
Making a third circuit wO'uld give back the value fl for f. Clearly f is a double-
valued function of z (Figure AA).
We have seen that it requires two circuits around the origin of the z-plane
to encircle the origin of the I-plane, thus giving rise to the double-valued-
ness off. On the other hand, starting with some point Zo (#- 0) in the z-plane
with well-defined argument 16 </J and following a closed contour C, one traces
16 More precisely, defined up to integral multiples of 4n.
z-plane I-plane
(a) (b)
Figure A.4 The function f(z) = ZI/2 (a) z-plane; (b) f-plane.
(a) (b)
Figure A.S Closed contour in one branch of the function fez) = ZI / 2 (a) z-plane;
(b) f-plane.
out a closed contour C' in the f-plane (Figure A.S). Thus, as long as one
does not encircle the origin one can treat f as a single-valued function.
The two values of ZI / 2 that occur for each value of z form two independent
sets, called branches of zl /2 (e.g., upper and lower zllz-plane). On crossing
the positive real z-axis one goes over from one branch of zl/2 into the other.
The positive real axis is called a branch line. If one analytically continues
one branch of the function zl i Z along a circle that enclosed the origin, one
ends up with the other branch. The region of encirclement then contains a
singularity of Z1 /2. Since the circle can be made arbitrarily small, it is clear
that the singularity is z = O. This is called a branch point of the function z1/2.
At a branch point, the multivalued function fez) has the same value for
all branches of z. Branch points always occur in pairs, with a branch line
joining a pair of branch points. In our example z = Cf) is the other branch
point. Any curve joining these two branch points can serve as a branch line
-the choice is a matter of definition and is usually dictated by convenience.
j. Riemann Surfaces
The theory of analytic functions outlined above for single-valued functions

can be extended to a wide class of multi valued functions using a geometrical
construct called a Reimann surface.
We shall describe the Riemann surface for the function fez) = Zl / Z. We
take two z-planes, call them R I and R 2, and define the branch line to be the
positive real axis. We "cut" R I along the branch line (a branch cut), or
actually" a little below" it; we do the same for R 2 .
We now join the lower lip of R I to the upper lip of R 2 , and the lower lip
of R z to the upper lip of R I . The surface thus obtained is the Riemann surface
for the function Zl / Z (Figure A.6).
We now define ¢ on R J to have the range 0 -;:;; ¢ < 2n, and ¢ on R z to
have the range 2n -;:;; ¢ < 4n. Thus on the entire Riemann surface 4> takes on
the values 0 -;:;; ¢ < 4n. Starting from ¢ = 0 in R I and encircling the branch
point, one reaches the branch cut. Crossing the branch cut takes us onto the
upper region, R z . Crossing the branch cut once more after encircling the
origin, one returns to R I . For z in R J [R2J one has fez) = fl(z) [fz(z)J,
Figure A.6 Riemannsurface for fez) = Z1 /2.
i.e., each Riemann sheet of the composite Riemann surface corresponds to a

single-valued branch of the composite multi valued function.
What we have achieved is the following: From a sequence of single-
valued functions [branches of the multivalued fez)] defined on one complex
z-plane, we have obtained one continuous single-valued function defined
on the Riemann surface. fez) is now analytic over the entire Riemann surface
except for the branch points, which are now to be treated as (isolated)
singularities.
A branch point is said to be of order n if the (multi valued) function is
returned to its original value after encircling the branch point at least n + 1
times. If this cannot be done for finite n, the branch point is said to be of
infinite order. Thus for the function zl /n, the branch points z = 0 (and
z = (0) 17 are of order n - 1. The Riemann surface has n sheets.
An example of an infinite-valued function is the function
fez) = In z
= In p + i(1) + 2nn); (A.6)
each encirclement of the branch point z = 0 increases the value of In z by
2ni. Accordingly, z = 0 is a branch point of infinite order, and the Riemann
surface has an infinity of sheets.
Problems
1. The square-well potential is given by
VCr) ={ °
-v
0
ifr < a.
ifr:$ a.
(a) Calculate the Born approximation of the scattering amplitude.
(b) Show that the exact expression for the lth partial-wave amplitude is given by
1 pjlka)j;(pa) - kNka)j/(pa)
7;(P) = iP kj;(ka)h/(pa) - pjlka)h;(pa) '
where k = J
p2 + 2m Vo (momentum inside the well) and j;(z) = dHdz, h;(z) =
dhidz.
(c) Show that at low energies the s-wave, I = 0, dominates so that T(p, 0) ~ To(p).
17 fez) is said to have a singularity at z = CIJ iff(l /z) has a singularity at z = o.

Problems 451
(d) Compare the correct low-energy amplitude with the Born approximation and
show that at low energies the Born approximation is good only for a shallow
well.
2. Consider the square-well potential of Problem 1 for the case of small energies.
(a) Obtain the scattering cross section and calculate the scattering length aa in
terms of a and Va in the limiting case
(b) Obtain the scattering cross section and the scattering length for the case that
Va is also small:
all
V ~--.
a2 2m
3. The repulsive square-well potential is given by
Va for r < a,
VCr) = {
o for r 2 a.
(a) Obtain an equation for the s-wave phase shift ba and express the value of ba in
terms of scattering momentum p and a for the limit Va ...... 00.
(b) Calculate the scattering cross section if the energy is so small that pa ~ 1.
(c) Compare the results of (a) and (b) with the results of Problem XVI.2 for the
the hard sphere.
(d) Study the p-wave phase shift b,= 1 for the repulsive square-well potential, in
particular for Va ...... 00, and show that it behaves like (pa)3. Compare it with
ba for the infinite repulsive square well Va ...... 00.
(e) Obtain the partial-wave amplitude for the infinite repulsive square well Va ...... 00
and for energies so small that (pa)" ~ 0 for n = 3, 4, ....
(f) Calculate the scattering length and the effective range for s-wave scattering for
the infinite repulsive square well Va ...... 00.
CHAPTER XVIII
Resonance Phenomena
Resonance phenomena constitute some of the most interesting and striking

features of scattering experiments. This chapter discusses in detail the con-
nection between quasistationary states and resonance phenomena, and cul-
minates in the derivation of the Breit - Wigner formula. In Section XVIII.2 the
concept of "time delay" is introduced and its relation to the phase shift
derived. Various formulations of causality are given in Section XVIII.3. In
Section XVIII.4 the causality condition is used to derive certain analyticity
properties of the S-matrix. These properties are discussed further in Section
XVIII.S. In Section XVIII.6, the central section of this chapter, the con-
nection between quasistationary states, defined by a large time delay, and
resonances, defined by characteristic structures in the cross section, is
derived. Section XVIII.7 describes the observable effects of virtual states.
Section XVIII.8 discusses the effect resonances have on the Argand diagram.
The actual appearance of resonances in experimental data when the effects
of the resonant phase shift, the nonresonant background, and the limited
resolution of the apparatus are taken into account is discussed in Section
XVIII.9.
XVIII.1 Introduction
In our study of the discrete energy spectra of atoms and molecules we treated
the excited states as if they were infinitely long-lived. When we discussed in
Section XI.4 the helium atom and its levels, which have the same energy
value as the physical system consisting of an He + ion and an electron, we
452
XVII!. 1 Introduction 453
ignored the presence of the He +-e - system and used an approximate descrip-
tion in which the helium atom was considered to be an isolated stationary
system, even though the helium-atom system can never really be isolated
from the He +-e - system with which it interacts. In this approximate descrip-
tion of excited states the transition processes could not be described. In the
more accurate description, which does not ignore transitions, excited states
have a finite lifetime. This lifetime is fairly long for the ordinary discrete
energy levels, but is not so long for the discrete energy levels that lie in the
continuous spectrum of the He-(He +-e-) system; the lifetimes of both sorts
of states are still, however, much longer than the time taken by the transition
processes.
The helium energy-loss experiment, whose result is depicted in Figure
XL4.1, is an inelastic scattering process in which a long-lived intermediate
state He* is produced, which subsequently decays either into a helium ion
and electron or into a helium atom and photon:
e- + He --4 e- I
+ He*
I ~He+ + e- (1.1)
He + y.
The lifetime of a singly excited He*(S), which has an energy level below the
first ionization threshold, is orders of magnitude larger than the lifetime of
the doubly excited He*(D).l This is due to the fact that the interaction VI, that
causes the transition
He*(D) --4 He+ + e-
is much stronger than the interaction VI that causes the transitions
He*(D) --4 He +y
and
He*(S) --4 He + y.
[The transitIOn He*(S) --4 He+ + e- is energetically impossible.] In the
approximation of Section XL4, in which both VI and VI, are ignored, the
energy operator is given by
H=K+V,
where K = Kelectron + K He with K He given by (XL 1.2), and V describes the
interaction between He and e -. Consequently both He*(S) and He*(D)
have infinite lifetimes. If VI is ignored but the stronger interaction VI, is not,
then He*(S) lives infinitely long, but He*(D) has a finite lifetime, which is
determined by VI,. If neither VI nor VI, is ignored, then both He*(S) and He*(D)
have finite lifetimes.
1 The width of the bumps in Figure XI.4.1 is mainly due to the resolution of the apparatus-
caused, for example, by the energy spread in the incident electron beam-and not a measure of
the lifetime.
454 XVIII Resonance Phenomena
a b
T c
Figure 1.1 Schematic diagram of a production process.
Systems that have finite lifetimes are called quasistationary or metastable

states, in distinction to the stationary or stable states, which are" infinitely"
long lived. Below in Section XVIII.6 we will justify the name resonances
for the quasistationary states. When to consider a state as a resonance and
when to consider it as stable is a matter of the accuracy required of the
description; the distinction between (relatively) stable states [particles like
He*(S)] and unstable states [resonances like He*(D)] is in principle quantita-
tive rather than qualitative.
There are two main types of experiments in which resonances can be
obtained: formation experiments and production experiments. The process
(Ll) is an example of a production experiment. These experiments are
characterized by processes of the form
a + T -+ b +R -+ b + c + d, (1.2)
and are depicted by diagrams like that of Figure Ll Production experiments
always involve inelastic scattering processes. Formation experiments, on
the other hand, consist of simpler processes
a+T-+R-+a'+T' (1.3)
and are illustrated by diagrams like that of Figure 1.2.
A beautiful example of a resonance obtained in a formation experiment in
atomic physics is the Schulz resonance in the elastic scattering process
(1 A)
If one measures the elastic scattering cross section as a function of the scatter-
ing energy (kinetic energy of the projectile e - with respect to the target He),
one observes that at the energy E = ER = 19.31 eV something unusual
happens. At this value the cross section changes violently, as shown in the
data of various elastic-collision experiments [cf. Figures 1.3, lA, and 1.5,
a a'
R
T T'
Figure 1.2 Schematic diagram of a formation process.

XVIII.l Introduction 455
lp
(a) 21 31 41 5111
lp
21 31 41 511
IS
21 31 41 511
21 lS
31 41 5111 ,
Helium
energy
levels
19 20 21 22 23 24 25
Electron energy (eV)
Figure 1.3 Typical current-energy plots for electrons in helium as observed by C. E.
Kuyatt, J. A. Simpson, and S. R. Mielczarek. [From Physical Review 138, A385 (1965),
with permission.]
which show the observed cross sections (Figures 1.4(a), 1.4(c), 1.5) as well as
the intensity of the transmitted electron current, i.e., the intensity of the
current of electrons that are not scattered (Figures 1.3, l.4(b), 1.4(d))]. We
shall show that the effect shown in these figures can be explained from the
hypothesis that He and e - form a long-lived compound that belongs to one of
the energy levels of the He- ion, as indicated by Equation (1.4).
A resonance is usually defined as a sharp structure in the cross section
together with a rapidly increasing phase shift going through n/2. In the
remainder of this chapter we shall see that in general the formation of a
quasistationary state is connected with these effects, and shall study the
characteristic properties of the S-matrix and phase shifts for these phenom-
ena. We shall in this chapter restrict ourselves to the case where only one
channel is open. This means that we are in an energy range below the first
inelastic threshold, where only the elastic channel
(1.5)
is open, so that the only reduced S-matrix element different from zero is
StCE A) = <'7AllstCE A)II'7A) = e 2io1 (EA)
(Recall that '7A are the quantum numbers that determine the internal state
of the a-Tsystem; we omit the label'7A on Sl and 6 1 ,)
Resonances in a multichannel system will be discussed in Chapter XX;
it will be shown there that with a few changes multichannel resonances may
be treated in much the same manner as the single-channel resonance case
studied in this chapter.
(b)
(c)
I
o
b 2·4
2()'S
Electron energy (e V)
Figure 1.4 (a) Variation of intensity of scattering of electrons at 72° by helium atoms
with electron energy, as observed by Schulz, Phys. Rev. Left. 13, 583 (1964). (b) Variation
with electron energy of the transmission of electrons through helium as observed by
Simpson, Phys. Rev. Left. 11, 158 (1964). (c) Variation with electron energy of the total
cross section for elastic scattering of electrons by helium atoms, as observed by Golden
and Bandel, Phys. Rev. 138, AI4 (1965) . • , Experimental points. (d) Variation with
electron energy of the transmission of electrons through helium as observed by Golden
and Nakano, Phys. Rev. 144, 71 (1966) .
10 per cent
~_ ~ of orrate
~---
-:i--
40°
1
9O~ I
1~
-v--- I
18.0 19.0 20.0 21.0

Electron energy (eV)
Figure 1.5 Variation with incident electron energy of the intensity of electrons elastically
scattered in helium at different angles of scattering. The intensity variations may be
judged from the lines shown on the right-hand side, which indicate 10 percent of the
total intensity near the resonance. From D. Andrick and H. Ehrhardt, Z. Phys. 192,99
(1966).
XVIII.2 Time Delay and Phase Shifts
The lifetime of a quasistationary state formed in a scattering experiment of

the form (1.3)2 is roughly measured by the time by which the projectile is
delayed in the scattering region as a consequence of the interaction. Let ~,
the interaction region around the target, be characterized by a radius R,
which is chosen large enough such that there is no interaction outside of 9l.
This is possible because the interactions encountered in practice do have a
finite range. We describe the state of the projectile-target system in the
Schr6dinger picture by the statistical operator W(t). We also imagine a
fictitious noninteracting system described by the operator Win(t). If H =
K + V is the total-energy operator, then the time development of W(t) is
generated by H, while that of Win(t) is generated by K [cf. Equation
(XIV.S.S)]. Since we shall consider only elastic scattering, with the target
remaining in the pure state WT = I17A) <17AI, we can ignore the factor WT in
Win = W~ <8> WT and consider Was referring to the projectile alone. W(t)
then describes the state as it develops when the target is present in the
center of ~, while Win(t) describes the state when there is no target in 9l.
2 To keep the discussion simple we assume the projectile mass remains unchanged, i.e.,
ilia' = ma = m.
The probability that the projectile is in the region f11 at the time t is given
by
let) = L d 3 x <xl W(t)lx) (2.1)
(cf. Sections 11.8 and 11.9), where Ix) are the generalized eigenvectors of the
projectile position operator with respect to the target. The time the projectile
spends in f11 is then
If there were no interaction between projectile and target, the time spent in
f11 would be
(2.3)
The time that the projectile is delayed by the interaction with the target is
therefore
tW) = T - Tin = L+: dt L d 3 x «x I W(t)1 X) - <X I Win(t)1 x»). (2.4)
This" time delay"3 may be positive or negative. If a quasistationary state is

formed it will be positive and large.
We now investigate the relationship of the time delay tW) with the phase
shifts and the S-matrix. We insert complete systems of K and H eigenvectors
into Equation (2.4) and use Equations (XIV.5.5) and (XIV.5.21) (which was
proved in Appendix XV. A) to obtain
tW) = Loooo dt L
d3 x ~ {<xla+)<a+le-itIlWeitllla'+)<d+lx)
_ <xla)<ale-itKwineitKla')<a'lx)}
= J:oo dt ~ e-it(Ea-Ea')<al Winla')
X L d 3 x{<xla+) <a'+ I x) - <xla) <a' I x)}. (2.5)
~ is the region within which the interaction takes place. The wave function
<x I a +) in this region depends strongly upon the particular nature of the
interaction. On the other hand, the wavefunction has a very general form
outside the region~, where the projectile moves freely, the effect of the inter-
action being expressed by a phase shift relative to the asymptotic form of the
free wave function (x Ia). Since we want to obtain statements about the time
3 Time delay has been introduced previously in a similar way by F. T. Smith, Phys. Ret'. 118,
349 (1960) based on E. P. Wigner, Phys. Rev. 98,145 (1955).
delay that are very generally true, we attempt to replace the integral over fll
by an integral outside of rJ£. To this end, we first show that the time delay
over all space (R -> r:I)) vanishes. For R -> 00 the last integral in (2.5) is
given by:
L d 3 x {<xla+)<a'+lx) - <xla)<a'lx)}
= L d3 x {<a'+lx)<xla+) - <a'lx)<xla)}
= <a'+la+) - <a'la), (2.6)

where we have used the relation
According to Equation (XIV.2.9a,b), the generalized eigenvectors of Hand

K have the same normalization. (This is proved in Appendix XV. A.) Thus
the right-hand side of (2.6) and consequently tboo ) vanish. We now split the
integral over the whole space in (2.6) into a sum of an integral over the region
9£ and an integral over the rest of the space outside the region rJ£, which we
denote by (00 - fll), and have
°
= L d3 x {<xla+)<a'+lx) - <xla)<a'lx)}
+ L-91 d x {<xla+)<a'+lx) -
3 <xla)<a'lx)}.
Inserting this into (2.5), the time delay over region rJ£ can then be written as
x J
(00-91)
d3 x {<xla)<a'lx) - <xla+)<a'+lx)}. (2.7)
Assuming spherical symmetry, i.e., [H, LJ = [K, LJ = 0, we may choose

angular-momentum bases {IEI131J+)} and {IEI1 3 1J)} for {Ia+)} and
{la)}.4 Then the time delay is given by
tf!) = f+ 00 dt I fp(E) dE peE') dE' e-it(E-E')<E 1131 winl E' 11 3)

- 00 113
4 Here 11 can be any set of additional quantum numbers, and we shall usually suppress the
label 11. If it is present then the position eigenket must also have this label: Ix, 11).
where p is the normalization function of (XIV.2.9a), and where the relations

<rB¢IEIl 3) = Yzh(B,¢)<rIE)b (2.9a)
<r B ¢ IE lln = YzI 3(B, ¢ )<r IE+)b (2,9b)
f dQ Yz I3(Q) Yz'I,(Q) = bll , bl313 (2.9c)
have been used. We can write (2,8) in the form
tb1t ) -cc foo dt'Lfp(E)dEP(E')dE'e-it(E-E')<EI13IWinIE'113)J, (2.10)

- 00 113
where
J == Loo r 2 dr {<rIE)I<rIE')t - <rIE+)I<rIE'+)n (2.11)
In Equation (2.11), which holds for all R larger than the radius of the inter-
action region f!Il, we choose R to lie in the asymptotic region, The integral J
can then be evaluated using the asymptotic forms (XVII.4.17) and (XVII.4.18)
of the exact and free radial wave functions. 5 Though derived under the as-
sumption of a Hamiltonian time development, these asymptotic forms exist
independently of H, depending as they do only on the S-matrix through the
phase shift bl • We have, therefore,
J = -
n
21 R
00
r 2 dr - -
1 , {.
r2pp
Sill ( pr - -In).
2
Sill (,
P r - -In)
2
- ei(6,(p)-6,(p')) sin (pr _I; + MP») sin (p'r _ '; + blP '»)}' (2.12)
Because the t-integration in (2.10) gives a factor beE - E'), one only needs
to evaluate J in the limit E' --+ E, i.e., P' --+ p. The evaluation of this integral
is an exercise in distribution theory, with the result
J(p --+ p') = .!.-\{dbdl(P) - (-lY.!.cos(2pR + blp» sin bl(P)}' (2.13)

np p P
[The calculation that gives (2.13) proceeds in the following way.

We rewrite the integrand in (2.12) as follows:
npp'J = .!. f+oo dr B(r - R){ei(P-P')TI _ e2i(6,(p)-6,(p'))]

2 -00
+ (_l)l[e i(p+p')r(e 2i6 ,(P) - 1) + e- i(p+p')r(e- 2i6 ,(P') - I)]}.
(2.14)
5 Only for I = 0 are the exact and free radial wave functions given by the trigonometric
function such that for I = 0 (2.12) holds for any value of R outside the interaction region and
one can choose for R the smallest possible value, the effective radius of the scatterer. The larger
the value of I the larger one has to choose the value of R in order to be in the asymptotic region.
XVIlI.2 Time Delay and Phase Shifts 461
The integral over the first term in the integrand is the Fourier
transform of the 8-function and may be calculated using Equation
(XV.2.2S):
first term = ! f+oo dr8(r - R)e i(P-P')Tl - e2i(d/(P)-d/(P'))]

2 -00
= 1.[1 _ e2i(d/(P)-d/(P'))]ei(P- p')R i (2.1S)

2 P _ p' + iO .
Using the relation 6 (VIII.2.S):
1 1 - . J( ') (2.16)
p-p _ I
'+'O=--,+m
p-p
p-p,
we rewrite (2.1S), noting that the contribution of the term containing

J(p - p') vanishes, since
We are left with
first term = HI - (2.17)

e2i (d/(P)-d/(P'))]e i(P- p')R _1_'-, .
p-p
Now let p -+ p'. Expanding the exponential in the square brackets,
we get
first term -+ H -2i(Jlp) - Jlp'»]ei(p-P')R _1_'-, -+ dJdlp) ,

p - p p
(2.18)
The integrals over the second and third terms in (2.14) are obtained
similarly, with the result
. [ ei(P+ p')R .
second and third terms = (-lY i .. ¥e+ 2 ,d/(P) - 1)
p + p' + 10
-~+~~ ]
_ i e . ·1.(e- 2id /(p') - 1) .
P + p' _ 10 2
(2.19)
Here there is no problem with taking the limit p' -+ p, and after
elementary simplification one obtains the second term on the right-
hand side of (2.13).]
6 Gel'fand and Shilov (1964), Vol. I.

If we insert J of (2.13) into (2.10), perform the integration over t [letting

the bound of the integral go to infinity and using (XIV.S.19)], and then inte-
grate over E', we obtain
t<l) = 2 L f p2(E) dE <E 1131 winl E 11 3)

113
X)2 [dd~) -(-lY~COS(2PR + c5[Cp)) sin c5[CP)). (2.20)
Since the expression in the square brackets does not depend upon 13 , we
can sum over that component of the angular momentum. Denoting
L <E 1131 winlE 113 ) = Wz(E) = Wz(p), (2.21)
13
which because of the normalization of Win fulfills

L fp(E) dE <E 1131 Winl E 1/ 3) = L fp(E) dE Wz(E) = 1, (2.22)
113 [
one can write (2.20) as

t<l) = ~ f pee) dE Wz(E)t<l)I(E), (2.23)
where for a nonrelativistic particle with mass m, the quantities p, E, and p

are related by [cf. (XIV.S.lI) and (XIV.S.18a)]
E = p2/2m, pee) = mp(E), (2.24)
and where we have defined
t<l)I(E) = 2p(E) )2 [dc5~~) - (-plY cos(2pR + c5z(p)) sin c5z(P)]
= 2m~ [d~~P) _ (-plY .![sin 2(pR + c5 l ) - sin 2PR]]. (2.2S)
The second term on the right-hand side of (2.2S) is of limited variation

and depends upon the radius R of the region fJf, which can be arbitrarily
chosen, the only restriction being that it be greater than the range of the
interaction. 7 We therefore average over R to obtain a quantity that does not
depend upon the irrelevant exact choice of the radius R and is characteristic
of the scattering process only. This average of t<l)1 is
[ (E) = 2 (E) ~ dc5z(p) = 2 ! dc5z(p) = 2 dc5z(E) (2.26)

tD P p2 dp m p dp dE .
The average time delay, which for short we again just call the time delay,
is thus
tD = ~ f pee) dE Wz(E)tb(E). (2.27)
7 Note that for I #- O. R in the above derivation has to be chosen in the asymptotic region.
The incident beam Win is usually not prepared with a definite value of the
orbital angular momentum I, but with a momentum pointing in a well-
defined direction. ~(E) then describes the weight of the Ith angular-momen-
tum component in the mixture Win. However, if one arranges that the
incident "beam" has a well defined angular momentum IA, then one has
p(E)~(E) = 611AFlE - E A ), (2.28)
where Fl)E - E A ) fulfills, because of (2.22), the condition
fdE Fl)E - E A ) = 1
and describes the energy distribution in the "beam" [cf. (XIV.5.35)]. The
incident beam is usually not monoenergetic. However, if it can be considered
monoenergetic, then [cf. (Xly'5.36)]
(2.29)
Inserting (2.29) and (2.28) into (2.27), one obtains for the time delay of this
particular" state" with angular momentum IA and energy EA
(2.30)
Thus the quantity (2.25) or (2.26) is the time delay occurring in the scat-
tering of a state with angular momentum I and energy E. And the time
delay in the state Win (2.27) is the weighted average of the time delays
tb(E) over all angular-momentum and energy values.
Restoring the dependence upon the additional quantum numbers 1],
the time delay of a "state" with quantum numbers 1]IE is given by:
I'/1(E) = 2 d67(E) = 2m d67(P) (2.31 )

tD dE p dp .
A monoenergetic beam describes a steady state for which time delay

really has no observable meaning. For a measurement of the time delay
one needs a beam of pulses of finite duration and, therefore, of nonzero energy
spread. 8 Then the quantity t D of (2.27) and also the time delay for a particular
angular-momentum value,
(2.32)
can-at least in a gedanken experiment-be measured as the difference in

time that a pulse of projectiles needs to pass through a region fJ1l with and
without the target present. Scattering experiments are, however, usually
not carried out in a way that allows the measurement of such a delay, and
the significance of (2.31) or (2.26) lies in its theoretical consequences, which
we shall discuss now.
8 Cf. the discussion in Section II. 9.

If during the scattering process for a particular set of values (l1IE) =

(IJRIRE R) a quasistationary state is formed, that is, if the projectile is tem-
porarily captured by the target, then for this set of values the delay time
t~(E) must be large. Thus the condition for a quasistationary state is
d61iE) has a sharp positive maximum at (IJR IRE R), (2.33)
which in particular means that
d 2 67J:(E) I _ 0, (2.34)
dE 2 -
E=ER
The main contribution in the expression (2.27) for the delay time then
comes from the term with these particular values (/RER)'
It may, of course, happen that for a particular scattering system all of the
M(E) are such functions of E that the d67(E)/dE do not have any maxima, and
therefore no quasistationary state exists. However, more often than not at
least some of the d67(E)/dE do have such isolated maxima, and we shall
investigate the consequences of such an occurrence below, in particular in
Section XVIII.6.
XVIII.3 Causality Conditions
The time T = T(M) which a projectile spends within a certain region :!It of
radius R around the target must always be positive. Thus the condition
T(M) ~ °
is obviously fulfilled. This condition is, through the definition of Tin (2.2),
connected with the von Kampen causality condition, which states:
L d 3 x <xl W(t)1 x) ~ °at any time t. (3.1)
(3.1) is also obviously fulfilled as it represents the probability to find the

projectile in the region:!lt. The strongest form of the condition (3.1) is obtained
if one chooses for R its smallest possible value, i.e, the effective radius of the
scatterer.
The time T(M) that the projectile spends in the interaction region can be
obtained according to (2.4) from
(3.2)
where tVf) is given by (2.23) and (2.25) and Tin(M) is given by (2.3) which can
be written:
Tin(M) = f+ 00 dt L fp(E) dE pCE') dE' e- it(E-E')<E 1131 Win IE' 11 3 )( _lin)

- 00 113
(3.3)
XVIII.3 Causality Conditions 465
with Jin given by
_J in = (R r2 dr (rIE>,(rIE'>t = ~ (R r2 dr j,(pr)j,(p'r). (3.4)

Jo n Jo
Equation (3.3) with (3.4) is obtained by the same arguments that led from
(2.4) to (2.10) with (2.11), except that in (3.4) we did not have to replace the
integral over the interaction region f!It by an integral over (00 - ~).
The integral in (3.4) is calculated using (XVII.3.9b) and (XVII.3.7a), the
result being
2
~ n (p
R2
+ p')(P -
( 1 . (R
p') pR 2sm p -"2
nl). ('R
sm p -
n(l -
2
1»)
1 .(
- p'R2 sm pR -
n(l 2- 1»).sm (p'R -"2nl)) . (3.5)
For the second equality in (3.5) the asymptotic form (XVII.3.6a) was used,
which is justified for R in the asymptotic region or in case I = 0 for R just
outside the interaction region. Using well-known relations between trigono-
metric functions, this becomes
- J in = -1 - 1 [-1 . (p - p ')R - -
- , sm Ysm
( --1, . (P + p')R] (3.6)
npp' p - p p+P
and for p' -+ p
_JIO ( -IY
. = - 1 ( R - --sin2pR . ) (3.7)
np2 2p
This expression is inserted into (3.3) after one has performed in (3.3) the
integration over t using S~ ~ eitx dt = 2nc5(x) and over E' which leads to
With the definition (2.21) and (2.24) one obtains
T in(9l) = ~ fpee) dE a;(E)T in(9l)'(E) (3.9)
where
T in (9l)'(E) -_ m ( pee)
2R (-1)'. 2 (E)R) (3.10)
- p(E)2 sm p .
Equations (3.9) and (3.10) have been written in complete analogy to (2.23)
and (2.25). The quantity T in(9l)I(E) is thus the time which a "state" with
angular momentum 1 and energy E spends in the region f!It if no interaction
takes place. The time T(9t) which the projectile spends in the region ~ when
the interaction is present is in analogy to (3.9) and (2.23) also written as
T(9t) = f
~ pee) dE W,(E)T(9t)I(E), (3.11)
where according to (3.2)

T(9t)I(E) = tC:)I(E) + T in (9t)I(E). (3.12)
Inserting (2.25) and (3.10) in (3.12), one obtains
T(91)I(E) = 2m! (db 1 + R) - (-lY ~ sin 2(pR + b1(p» (3.13)

p dp p
T(91)I(E) is the time that the "part" of the projectile with energy E and angular
momentum 1spends in the interaction region.
The condition (3.1) must be fulfilled for any arbitrary state of the projectile,
i.e., for any set of positive functions W,(E). Consequently we conclude
from (3.11) and (3.1) that
T(91)I(E) ~ 0 for alII and E. (3.14)
This condition (3.14) can be rewritten using (3.13):
db 1 ~ -R
~
+(_ ly sin2(b1
~
+ pR) ~ -(R +~).
~
(3.15)
In the form (3.15) the condition is called Wigner's causality inequality. The
smaller R the stronger is the condition (3.15); but only for I = 0 can one
choose in our above derivation for R the effective radius of the scatterer.
Wigner's causal inequality states that the phase shift cannot decrease faster
than at a certain rate. Thus if the phase changes rapidly, then it must be
increasing. We shall make use of this fact below.
The condition (3.1) is so obviously fulfilled that one may wonder why the
conditions (3.1) and (3.15) were given the name causality conditions. The
connection with causality becomes most apparent if one rewrites (3.14) into
yet another form using (3.12) with (3.10):
tC:)I(E) ~ _ T in(91)I(E) = _ [m. 2R _ (_1)1 m sin 2PR] . (3.16)

p p p
The first term on the right-hand side of (3.16) and of (3.10) is the free time of
flight that a classical projectile with velocity p/m would need to traverse the
region of diameter 2R. The second term represents the quantum effects;
it is negligible for sufficiently fast particles, pR ~ 1, and arises from the fact
that quantum particles cannot be localized within dimensions smaller than
a de Broglie wavelength.
If the region of diameter 2R is the interaction region in which the projectile
interacts with the target, then-depending upon the kind of interaction-the
projectile may be delayed (tD > 0) or hastened (tD < 0) in it. The projectile
may be delayed an arbitrarily long time, and we shall discuss this phenomena
XVIII.4 Causality and Analyticity 467
in the subsequent sections. However, it cannot be advanced an arbitrarily

large time, because according to causality outgoing projectiles cannot appear
before the incoming projectiles have reached the scattering region.
Classically, the maximal time advance occurs if the projectile has hardly
entered the interaction region when it is repelled at the surface, and leaves
the interaction region [Jl at a time m2R/p earlier than a particle that passes
through such a region without interaction-i.e., its time delay is -m2R/p.
A more negative time delay is not possible, because then the outgoing pro-
jectile would have to leave the surface of the interaction region before the
incoming projectile had reached it. For quantum systems these arguments
have to be corrected by the term (m/p)(sin 2pR)/p, and then lead to (3.16)
as a mathematical formulation of the above statement of causality.
XVIII.4 Causality and Analyticity
In chapter XVI we considered the lth partial T-matrix elements <1Jb II 1I(E) II 1JA)
and the lth partial S-matrix elements <1JbIISlE)II1JA); in particular we con-
sidered the lth partial T- and S-matrix elements for elastic scattering,
1I(E) = 1I(P) and SlE) = Slp), where the momentum p and energy E
of the projectile of mass m are connected by E = p2/2m. If the interaction is
described by a Hamiltonian H = K + V, then T and S are connected
to the interaction Hamiltonian V by expressions like (XIV.S.41), but as
mentioned above (e.g., in Section XIV.4), the notions of S- and 1'-matrix
make sense also if there exists no Hamiltonian time development. Then
the S- or T-matrix is the ultimate basis for the description of all the in-
formation required to compute observable quantities, such as cross sections
[(XVI. 1.26), (XVI. 1.29), (XVI.1.30)]. The properties of the S-matrix then
follow from general physical principles believed to be satisfied by any inter-
action. One class of general physical principles are symmetry principles,
and we have, in fact, already made use of such a symmetry principle, namely
the rotational symmetry (XVI. 1.1), when we derived that the S-matrix
elements are independent of 13 , The general physical principle that we shall
utilize now is the causality condition discussed in the preceding section.
The causality condition leads to analytic properties of the S-matrix elements
when the energy and momentum of the projectile are extended to complex
values. 9 When the interaction is described in terms of a potential function
VCr), then this kind of analyticity of the S-matrix can be derived from the
properties ofthe potential function. 10 Following the spirit ofthe presentation
in this book, we shall not start from a potential function, but establish some
analyticity of Sl(P) as a consequence of causality. We shall not give an exten-
sive discussion of this subject, especially as not all questions on this subject
9 One obtains dispersion relations when these analytic properties are expressed in terms of
integral relations between different matrix elements for real values of the variables.
10 Taylor (1972), Chapter 12.
have been answered yet, and content ourselves in this section with the formu-
lation of the analyticity properties of SI(P) that will be needed in the sub-
sequent sections in connection with resonance phenomena.
The probability of finding the projectile at an arbitrary time t somewhere in
space is unity:
1= f+oo d x <xIW(t)lx) = r d x <xl W(t)1 x) + r d3x <xIW(t)lx).

3 3
-00 J. J(oo-~
(4.1)
Consequently the van Kampen causality condition (3.1) can be reformulated
to read
r
Joo-.'Jf)
d3x <xIW(t)lx) :$; 1 at any time t. (4.2)
Inserting a complete system of basis vectors IE lit) and using (XIV.S.21)

and (2.9a,b,c), one obtains for (4.2) [by the same calculation that led from
(2A) to (2.7) except for the change from the integral over f!Jt to the integral
over (00 - f!Jt)]
L SSP(E) dE pee') dE' e- il(E-E')<E 1131 winl E' 11 3 )

113
x r:R <rIE+)I<E'+lr)lr 2 dr:$; 1.

OO
(4.3)
In the region (00 - f!Jt), where no interaction takes place and the projectiles
move freely, the wave function <rl E+)I = <rl P+)I has the general asymptotic
form (XVIIA.17), in which the effect of the interaction is expressed in terms
of the S-matrix Slp) = e2i6 /(p):
<rlp+) '" g_~_(ei(pr-I1tj2)S (p) - e- i(pr-I1t/2». (4A)

I .yn21pr I
Inserting this into the integral over r in (4.3), one obtains for R from the
f::
asymptotic region
integral = r 2 drO(r - R)<rlp+)I<rlp'+)f
= ~4;p,e-i(p-p')R{S/R(P)SMP') t+: drO(r - R)ei(P-p'Hr-R)
+ J +OO
drO(r - R)e-i(P-p'Hr-R)
f::
-00
-e- il1t S/R(p) dr OCr - R)ei(p+p'Hr-R)
-ei/1tSMp') t+: dr O(r - R)e-i(p+p'Hr-R)} ,

XVlI1.4 Causality and Analyticity 469
where we have defined l l

(4.5)
The Fourier transform of the 8-function has been calculated in (XV.2.25 + );
using that result one obtains
. _ 2 e -i(p-p')R { *, .
I
.
I
mtegral - -n 4 ' SIR(P)S/R(p) , + '0 +, + '0
pp p - pip - P I
+ (-lY(SfR(P') p +: _ iO - SIR(P) p + p~ + iO)}'

U sing the relation between the distributions, (2.16) = (VIII.2.5), this can
be rewritten
integral = - 2
1 , e-i(p-p')R{(SIR(P)SiR,(P') _l)~i_,
npp p- p
+( -lY(Si'R(p') - SIR(P» P ~ P'
+ nc5(p - P')(SIR(P)SfR(P') + 1)
+ n( _l)Ic5(p + p')(SfR(p') + SIR(P»}
Inserting this integral into (4.3), one obtains, using p dE = p2 dp,
L ('Xl roo p2 dp p,2 dp' e-it(E-E')<p 1131 winlp' 11 3 ) _1_, e-i(p-p')R

113 Jo Jo 2npp
x {(1 - SIR(P)Si'R(P'» i(p ~ p') -( -lY(SfR(p') - SIR(P)) i(pl p')

+ nc5(p - P')(SIR(P)Si'R(P') + 1) + n( -1)1c5(P + p')(Si'R(p') + SIR(P»} ~ 1.
(4.6)
Performing the integral over p and p' in the last term gives a zero contribution
11 The physical meaning of the exponential factor e -l;pR in

S,(p) = e2 ;6 ,(p) = e1 i(-pR)S'R(P)
is that it represents the phase advance corresponding to a path difference from the surface of the
interaction region to the center and back. According to the description of Figure XVII.4.2,
a repulsive potential gives a negative phase shift. An impenetrable sphere of radius R has a
phase shift of - pR, because the wave is reflected at the surface of the sphere, as compared with
the one going through the center, cf. Problem XVII.3. Correspondingly an outgoing signal can
appear up to 2R/v = 2Rm/p earlier than would have been possible in the absence of the scatterer.
S,R(P) then describes the deviation of the effect of the scatterer from the effect of an impenetrable
sphere with the radius of the interaction region.
because of b(p + p'). After integrating the next to the last term over p' and
using the unitary condition S/R(p)SiR(P) = 1, one obtains the contribution
L fd P p2<p 113lw in lp [[3) = 1.

II,
For t = 0 the above inequality (4.6) therefore becomes
I roo roo p2 dpp,2dp'<p1l3IW in lp'1l3)_1_,e- i(P-P')R

Jo Jo
!l3 2npp
x (SIR(P)SiR(P') - 1).( 1 ')

I P- P
Z L roo roo p2 dp p'2 dp' <p 1131 winl p' 1/3) _1_, e- i(p- p')R( _1)1
!l3 Jo Jo 2npp
x (S/R(p) - SiR(P'» i(p ~ p') (4.7)
As this has to be fulfilled for any state Win, it follows that a corresponding
inequality must hold for any term in the sum I1l3
separably and for any
physically permitted momentum distribution function.
One can write (4.7) in a different form by introducing for Win the momen-
tum wave function A~3(p) [in analogy to the f(p - p') of (II.9.1p)] by
A~3(p)A~3*(p') = pe - ipR<p 1/31 Win Ip' 11 3 )e ip 'Rp'. (4.8)
Thus one obtains from (4.7)
OCfOO 1
f o 0 dp dp' A~3(p)A~3*(p')(S/R(p)SiR(p') - 1) -.(----,-)-
lP-p
f foo dp dp'( -lYA~3(p)A~3*(p')(SIR(P) -

zoo OO
S'l'R(p'» i(p +1 p') (4.9)
for every value of I and 13 and for any well-behaved function (i.e., for any
element of the Schwartz space), A ~3(p).
From the form (4.9) of the van Kampen causal inequality it follows, by a
purely technical proof that we shall not reproduce here :12
S/(P) has an analytic continuation without singularities in the first

quadrant 0 ::;; arg p ::;; n/2 - b((j > 0) of the complex p plane. (4.10)
In fact, one can prove more about the analyticity property of SzCp) than just
this, and in Section XVIII.5 below we shall summarize some results. It is,
however, only (4.10) that we shall need for our derivation of the Breit-Wigner
formula in Section XVIII.6. We wish to emphasize that (4.10) holds for any
value that one chooses for R, as long as it is finite.
12 N. G. van Kampen, Phys. Rev. 91,1267 (1953), Section II.

XVIII.5 Brief Description of the Analyticity Properties of the

S-Matrix
Though a derivation of the following result on the analyticity properties of

Sz(p) exceeds the scope of this book, and we shall also not make explicit
use of these properties in what follows, the knowledge of these properties
will greatly enhance the picture of the S-matrix. We shall, therefore, state
these analyticity properties here with a few connecting remarks. 13 These
properties are independent of the specific form of the interaction and follow
from very general principles-essentially from the causality condition in
the form (4.9).
In order to obtain the properties of SI(P) beyond the first quadrant, we
have to continue Slp) across the imaginary axis and into the lower half
plane. In order to do the first we have to know the properties of SI(P) on the
positive imaginary axis. We know already from the discussions in Section
XVII.5 that bound states correspond to poles on the positive imaginary axis,
so that we cannot expect SI(P) to be analytic there. Figure 5.1 shows these
poles and zeros, which are, of course, also possible on the real axis.
As a further consequence of the causality condition (4.9) one can prove that
I Rep
1m (-1) S/R(p) ::;; - , 1m SIR(P) ::;; 1 in 0::;; arg P < nl2 (5.1)
Imp
and
I SIR(P)1 is bounded in the first quadrant. 14 (5.2)
This is not sufficient to obtain information about the nature of the singular-
ities on the imaginary axis, except to conclude that if these are poles, they
cannot be of higher than the first order.
In order to conclude anything more, one has to make an assumption in
addition to the causality condition. If the interaction is described by a
Hamiltonian H, this assumption amounts to the requirement that H be a
semi bounded operator, i.e., there is a lowest energy eigenvalue B and the
13 For more on this subject, see Nussenzveig (1972).

14 This is a very weak condition for large values of R, because it allows S/(p) to vary rapidly
on the real axis also in the absence of any poles or zeros. But in the above derivation of the
causality condition (4.9) R has to be chosen from the asymptotic region because (4.4) is the
asymptotic form for the radial wave function outside the interaction region. The smaller the
°
value of I, the smaller can one choose the value of r for which the asymptotic form (4.4) is valid.
°
But only for the case 1 = is (4.4) identical with the radial wave function outside the inter-
action region. Therefore, only for 1 = can one choose for R the smallest possible value namely
the effective radius of the scatterer. In order to obtain the strongest result, i.e. the above state-
ments for R equal to the interaction radius of the scatterer, also for 1 > 0, one has to take instead
of the asymptotic form (4.4) the exact expression for the radial wave function for r outside the
interaction region:
(4.4')
where h,(pr) and hi(pr) are the spherical Hankel functions (XVII.2.5). The mathematical proof
that (5.1) and (5.2) follow also if one takes (4.4') instead of (4.4) has not yet been given.
binding energy cannot be infinite. In the case where one does not have a
Hamiltonian time development and the bound states are described by
singularities on the imaginary p-axis, one would assume this to translate
into the statement that there are no singularities of Slp) on the imaginary
p-axis above the value iK, where K = + J2mB.
It has in fact been shown 15 that from a precise formulation of the finiteness
of the binding energy it follows that
Rep .
1m (-lYS1R(p) ~ - ~- III the first quadrant, for K = 0 (5.3)
Imp
and
1m SIR(P) --> 0 for Re p --> 0+, for any K ~ O. (5.4)
Consequently
Slp) is real on the positive imaginary axis except between 0 and iK (5.5)
For scattering states the physical values of the momentum p must be
positive; however, p occurs quadratically in the Schr6dinger equation and
also in the Lippmann-Schwinger equation. Consequently the equations are
invariant with respect to the change of sign of p, and therefore, if p is replaced
by - p in the radial wave function, the resulting function must again be a
solution of these equations. Replacing p by - pin (XVII.4.17d) gives
-i(7t/2)(1+ 1) 1
<rl- p+> ~ e (-lYS/-p)_(-(-lYe- ipr + Sil(-p)e ipr ).
~ pr
(5.6)
As <r 1p + >and <r 1- p + >describe the same physical content, the right-hand
side of (XVII.4.17d) and (5.6) must both describe an incoming spherical wave
and an outgoing spherical wave modified by the scattering matrix. We
therefore require that the right-hand side of (5.6) agrees with the right-
hand side of (XVI1.4.l7d) except for an irrelevant phase factor « -l)SI( - p».
This leads to the condition that for negative values of p the S matrix element
Sl be given by
SI-I( - p) = SI(P) (5.7a)
or
Sl(-P) = SI-l(p). (5.7b)
Using unitarity,
Slp)S((p) = 1, (5.S)
one obtains
Sl-p) = S((p) for real p. (5.7c)
This is called the symmetry relation for the lth partial S-matrix element.
15 N. G. van Kampen, Physica (Utrecht), 20, lIS (1954).

Bound state /
po les <.......
> Virtu al
state zero
x X
Re onance
poles
Figure 5.1 Some poles (indicated by X) and zeros (indicated by 0) of the S-matrix
element in the p-plane.
The Ith partial S-matrix element is usually considered as a function of

E = p2/2m, SlE). In the mapping from p to E the complex p-plane (see
°
°
Figure 5.1) is mapped onto a two-sheeted Riemann surface [cf. XVII.Ai
and XVII.AjJ with the branch cut from to 00. The upper half p-plane
1m p > corresponds to the first sheet; the first quadrant of the p-plane
corresponds to the upper half of the first sheet. The physically meaningful
values of p in scattering states, p real, p ~ 0, correspond to the upper rim
of the first sheet. The first sheet is called the "physical" sheet (Figure 5.2). If
one continues through the cut one comes to the second sheet, called the
"unphysical" sheet, which corresponds to the lower half of the p-plane.
SlE) is a function on the two-sheeted Riemann surface. The bound-state
poles of Sl(P) on the positive imaginary p-axis correspond to poles on the
negative real E-axis of the" physical" sheet. Statement (4.10) then means
that
Sl(E) is analytic in the upper half plane oj the physical sheet. (5.9)
E-plane (physical sheet)
"",cut
--~--~----;-----------~
Bound state poles
Figure 5.2 Poles and cut in the first C physical") sheet of the energy plane.
The relation (S.7c) for Sl considered as function of E is then stated as

SlE - if) = Sf(E + if), E> O. (S.10)
From the Schwarz reflection principle (XVII.Ah) and condition (S.5)
follows that SI(E) can be continued into the lower half energy plane across
the negative real energy axis and is given there by
(S.11 )
Since the lower half energy plane corresponds to the second quadrant of
the momentum plane and E* corresponds to - p*, (S.11) leads to
Slp) = st( -p*) (S.12)
as an extension of the symmetry relation (S.7c) to complex values of p. Thus,
at points symmetrically placed with respect to the imaginary axis, SI(P)
takes on complex conjugate values. As it has no singularity in the first
quadrant, it has also no singularity in the second quadrant. If SI(P) has a
zero at PI' as shown in Figure 5.3, then it will also have a zero at - pi.
One can now continue SI(P) into the lower half p-plane (Figure 5.3), by
extending the relation (S.7a) to complex values of p. For every value in the
second quadrant (5.7a) defines a value of SI(P) in the fourth quadrant, except
at zeros in the second quadrant which produce poles in the fourth quadrant.
In the same way (S.7a) defines a function analytic in the third quadrant
except for poles coming from zeros in the first quadrant. Thus, Slp) is in
general not analytic in the lower half plane and if it has zeros above the
positive p-axis it will have poles below the p-axis, the resonance poles of
Figure 5.1 and their counterparts in the third quadrant.
In this way we have seen that Slp) is a meromorphic function in the whole
p-plane that is analytic in the upper half p-plane except perhaps on the
imaginary axis. The poles in the lower half plane occur in pairs, symmetrically
distributed with respect to the negative imaginary axis, except for those on
the negative imaginary axis, corresponding to zeros on the positive imaginary
p-plane
o o
op,
- p, x x p1
x x
Figure 5.3 Relation between zeros (indicated by 0) and poles.
E-plane (unphysical sheet)
CUi
x X
Virlu al sta te po les Reso na nce po les
Figure 5.4 Positions of possible resonance poles, virtual-state poles and cuts (capture
state poles not shown) in the second (" unphysical ") sheet of the energy-plane.
axis. [The number of poles can be infinite, but they cannot have an accumu-
lation point for finite values of p, as an analytic function (in the upper half
plane) cannot have a finite accumulation point of zeros.] The singularities
on the imaginary axis can be shown to lie within the interval ( - iK, + iK),
where B = K2j2m is the maximal binding energy. Nothing can be derived
from causality about the behavior of S/(p) in the neighborhood of this interval.
It is, however, believed that for an interaction that is effective only in a
region of finite size, one has poles, which must then be simple. 16
If we translate these properties of S/(p) into statements about SlE) in the
energy plane, we can summarize our results, which are principally a con-
sequence of the causality condition, in the following way (Figures 5.2 and 5.4):
SlE) is a meromorphic function on the two-sheeted Riemann surface with a
branch point at E = 0 and a cut from 0 to 00. The physical values of E in
collision processes lie on the upper edge of the cut on the "physical sheet."
Bound-state poles lie on the negative real axis of the physical sheet, and
except for these bound-state poles SlE) is an analytic function on the physical
sheet. Further poles (of any order) may lie on the second, "unphysical"
sheet, coming from possible zeros on the first sheet. Poles at various locations
on the "unphysical" sheet have various physical interpretations. Poles on
the negative real axis of the unphysical sheet [coming from zeros of Slp)
on the positive imaginary axis] are called virtual-state poles. A virtual state
is one that would be bound if the interaction were more attractive and a
virtual state close to threshold causes a large cross section at low energy;
we will discuss such states in Section XVIII.7. Poles of SlE) on the un-
physical sheet, if they are close to the positive real axis, are of particular
16 If the interaction is described by a potential VCr) that decreases faster than any exponential
for r -+ 00, then S,(P) is meromorphic in the whole p-plane and analytic in the upper half plane
except for a finite number of simple poles. On the other hand, for a Yukawa-type potential
VCr) = J: (J(/l)e-w/r d/l, m > 0, one has, in addition to possible poles of S,(p), also cuts along
the imaginary axis from im/2 to ioo and from -im/2 to -ioo.
importance. They are called resonance poles or Siegert poles, and we shall
study them in the remainder of this book. Each pole in the second sheet
below the real axis has a counterpart in the second sheet above the real axis.
As we shall see in Chapter XXI, the pole below the real axis describes a
decaying state and the pole above the real axis describes a forming (capture)
state.
XVIII.6 Resonance Scattering-Breit-Wigner Formula for

Elastic Scattering
Experimentally, resonances are usually associated with a sharp variation of

the cross section as a function of energy. If the elastic or inelastic cross sections
exhibit sharp maxima or minima, one says that a resonance has occurred.
From (XVI.2.9) for the elastic Ith partial cross section one sees that a maxi-
mum occurs either at the energies for which c5[(E) = nl2 (modulo n) or for
which c5[(E) has a maximum. The latter possibility, however, cannot lead to
a sharp maximum, because according to the causality condition (3.15), the
phase shift can never decrease with a steep slope. Thus a sharp maximum in
the lth partial cross section occurs at those value of E for which c5lE) = nl2
(modulo n); and the more rapidly c5lE) increases by n, the sharper is the
maximum. Similarly, a minimum for the elastic Ith partial cross section
occurs for c5lE) = n (modulo n). The connection between the change in
cross sections and the change in phase shifts for these two cases and for some
intermediate cases is shown in Figure 6.1. Note that as a consequence of the
causality condition (3.15), c5[(E) must be mainly increasing; if it decreases,
this decrease can be only very gradually. In every case a sharp structure in the
cross section alE) is connected with a sharp increase of the phase shift
c5lE) by n.
We therefore want to take as the preliminary definition of a resonance of
angular momentum I at the energy Eo that
(jl(E) increases rapidly by approximately n when E passes through Eo. (6.1)
If there is a sharp increase of one (j[ by n around Eo, and if all other phase
shifts are constant or slowly varying around Eo, then rapid variations will
also show up in the total cross section a = I a[ or the differential cross
section at certain angles. However, not all structures in the cross section
should be ascribed to resonances. Resonances can usually be distinguished
from nonresonant phenomena in the cross sections by the fact that resonances
appear only in the single partial wave, while nonresonant phenomena are
the result of cooperative contributions from many partial waves.
We now want to explain the association of resonances and quasistationary
states that we have already alluded to in the introduction (Section XVIII.1).
According to our considerations in Section XVIII.2, a quasistationary
state occurs at those values of E = Eo for which d(j[(E)ldE has a sharp
271
Ii
3-
2
II
O,(P)
71
sin' O,(P)
Eli' Eli" £I"'

o E\," £<0"
(a) (a') (b) (c) (d)
E •
Figure 6.1 Phase shifts and resonance profiles.
maximum. We therefore take as the definition of a quasistationary state of

angular momentum I at the energy Eo that
~2bl(~ I = 0, d3 b,\E) I < 0; ~ Idb,(E) I ~ Id3 b,\E) I . (6.2)

dE E=Eo dE E=Eo Eo dE Eo dE Eo
Whereas the first two conditions of (6.2) state that Eo is a maximum of

db,(E)/dE, the last condition of (6.2) states that this maximum is to be very
sharp. The first part of the conditions (6.2) is already incorporated in the
drawings of the phase shifts in Figure 6.1, where the point Eo around which
b,(E) changes rapidly is a point of inflection. However, it is not clear that
the two phenomena characterized by (6.1) and (6.2) are related. We shall
show in this section that as a consequence of very general physical as-
sumptions (causality), these two phenomena are indeed connected: that a
quasistationary state leads to a resonance and that a resonance leads to a
quasistationary state. We will see that (6.2) leads to a pole in the lth partial
S-matrix below and near to the positive real E-axis. And we shall obtain
from (6.2) a representation of S,(E), valid in the neighborhood of Eo, from
which (6.1) follows immediately.17 The notion of resonance poles was

introduced in Section XVIII.5, when we described the analyticity property
of the S-matrix. We argued there that SI(E) can have poles anywhere in the
second sheet, which came in pairs unless they lie on the negative real axis,
and we mentioned that they are called resonance poles if they are close to the
positive real axis. Here we show that they appear as a consequence of the
existence of a quasistationary state, (6.2), and provide therewith a physical
explanation of their existence.
To start our derivation we introduce the function
(6.3)
which as a consequence of (6.2) has a sharp minimum at E = Eo, so that

!' = !'(E)E=Eo = 0 and f" = f"(Eh=Eo is large. More precisely, a simple
calculation shows that
f ff = _ (db 1 (E
dE 0
») -2 d 3 b1 (E )
dE3 0, (6.4)
and consequently,
(6.5)
from the last two inequalities (6.2).

We want to expand feE) in a Taylor series around E = Eo:
(6.6)
where the" remainder" term is

fm 3 j<n) n
F(E-E o)=3T(E-E o) + ... +~(E-Eo) +... (6.7)
The Taylor series converges if for all n larger than a certain value N
j<n+l) < j<n) n + 1 (6.8)

(E - Eo)'
certainly not too stringent a condition.

We shall first assume that F(E - Eo) is negligibly small [F(E - Eo) = 0]
in a certain energy range around Eo ; this would be the case if b\4) and the
higher derivatives of the phase shifts at Eo could be neglected. In a second
step we will discuss the corrections that arise from a small F(E - Eo).
17 This derivation is based on Goldberger and Watson (1964), Section 8.5.

With F(E - Eo) = 0 it follows from (6.3) and (6.6) that
£5 (E) -
1-
fI + dE
1"/2(E - Eo)2
= JI~" arctan (J2~/;~) + Y;

= -
(2
-Vff (E - Eo)
arccot J2f11" + Yl (6.9)
(2
£5lE) = - -VJrarctan (J2f11")
E _ Eo + Yl' (6.10)
where Y; and Yl are arbitrary integration constants,
y, - Y; = J(2If1")n/2.
Equation (6.10) suggests the introduction of the two new parameters:
r=Jf,= -~
2(W»)3
(6.11)
and
I
2£5j1)
r=2 -=2
I" - W)' (6.12)
r is dimensionless, and :C has the dimension of energy. As a consequence of

(6.5) (which followed from the requirement that there exist a quasistationary
state at Eo), r fulfills the condition
r ~ 2E o . (6.13)
Inserting (6.11) and (6.12) into (6.10), one obtains for the Ith partial s-
matrix element
(6.14)
The phase shift (6.10) is written as a sum of two terms:
(6.15)
where £5!R)(E) is a rapidly varying function due to (6.13):
£5(R)(E) = r arctan =-r_/2_ (6.16)

, Eo - E'
whose value changes by almost rn when E passes through Eo. y, is a constant

under the assumption F(E - Eo) = O.
h,
H____ _
~.{ - - --
E
r r
Eo - 2" Eo Eo + 2"
Figure 6.2 Change of phase shift at a quasistationary state.
If F(E - Eo) is not zero, then one obtains instead of (6.9):
c5lE) =
Jf + dE
(f"/2)(E - Eo)2 + F(E - Eo) =
Jf + dE
(f"/2)(E - EO)2
-
J [f + (f"/2)(E
dE F(E - Eo)
- Eo?][f + f"/2(E - Eo)2 + F]' (6.17)
Writing c5lE) again in the form (6.15), we obtain for YI
YI = ylE) = - J [f + (f"/2)(E
dE F(E - Eo)
- Eo)2][f + (f"/2)(E - Eo)2 + F]'
(6.18)
which for small F(E - Eo) is a slowly varying function of E.
Thus the phase shift in the neighborhood of a quasistationary state defined
by the requirements (6.2) is the sum of a term c5\R)(E), which changes rapidly
by almost nr, and a slowly varying term YI(E):
(6.19)
This is depicted in Figure 6.2. The quantity YI(E) is called the background
phase shift, or also the potential part 18 of the phase shift, as opposed to c5\R)(E),
which is called the resonant part. Thus, if r = 1, we have derived the con-
dition (6.1) as a consequence of the condition (6.2). In particular, if the
background phase shift is zero, we have the situation depicted in Figure 6.1(a).
We should now, therefore, ask what values the parameter r can take. In
order to answer this question, we use the analyticity property (4.1 0).19
The function
r/2 i
arctan E = -2 [In(E - Eo + ir/2) - In(E - Eo - ir/2)] (6.20)
Eo -
18 If the interaction is described by a potential, then Yl comes from the long-range component
of this potential, whereas /if comes from a short-range attractive component.
19 For the present discussion not all of the analyticity (4.10) is required; one just needs
that StCE) and, therewith, 7;(E) are analytic in a domain above the real positive energy axis
that contains the point E = Eo + ir/2, a very weak condition indeed.
has a branch point of infinite order at E = Eo + ir/2 (and also one at

E = Eo - if/2).20 If E moves from the real axis around this branch point
and returns to the initial point on the real axis, always staying in the domain
of analyticity, then In(E - Eo - ir/2) increases by 2ni, and arctan[(f/2)/
(Eo - E)] by n. Thus moving along this closed contour around the point
Eo + if/2 results, according to (6.14), in the change
SlE) ~ SlE)e 27tir . (6.21)
On the other hand, since SlE) is analytic in this domain, its value should
not change. This is only possible if
r = 0, 1,2, .... (6.22)
We shall now discuss the simplest nontrivial case, r = 1. From (6.14) and
(6.15) one then obtains
SlE) = Eo + ir/2 - E eiZ;, =

Eo - ir/2 - E
(1 + Eo - if
ir/2 - E
) ei2J"1
(6.23)
This shows that the S-matrix has a simple pole at the value E = Eo - ir/2,
which is called a Siegert pole. For r = - 1 we obtain another representation
of the S-matrix
S (E) = Eo - if/2 - E.e,2y, = ( I + - if ) e''Z'i, (623')

I Eo + if/2 - E Eo + ir/2 - E .
which has a pole at E = Eo + if/2. These two poles are one pair of the
resonance poles mentioned in Section XVIII.5. There we showed that
resonance poles come in pairs; (6.23) is the representation of the S-matrix
most suitable for energies in the neighborhood of the pole below the real
energy axis and (6.23') is the representation most suitable for energies in the
neighborhood of the pole above the real axis.
We shall at the moment consider the case r = 1. From (6.16) follows:
R f/2 (6.24)
tan bl (E) = Eo _ E
The S-matrix element (6.23) is then written
SI(E) = 1 + 2ipTI(E) = ei2Y '(1 + 2ipTl R )(E)), (6.25)
where one has defined the resonant partial-wave amplitude TlR)(E) in such
a way that it agrees with the partial-wave amplitude 7;(E) if the background
can be ignored (Yl = 0). From comparison of (6.25) and (6.23) one then
obtains for the resonant partial-wave amplitude
(R) _ 1 r/2
Tl (E) - pEo _ ir/2 - E (6.26)
20 Cf. Appendix XVII.A.

In conclusion we have shown the following: if a quasistationary state

with angular momentum 1 is formed at the energy Eo, then the phase shift
has the resonant behavior specified by (6.1), and-if the background can
be neglected-the partial-wave amplitude is given for the simplest case
r = 1 by (6.26).
The Ith partial cross section that one obtains from (6.26), which is therefore
the pure resonance cross section (any background being neglected), is then
given [using (XVI. 1.38)] by
R 4n (r/2)2
(J/ (E) = p2 (21 + 1) (Eo _ E)2 + (r/2)2· (6.27)
This function of E is shown in Figure 6.3 together with the corresponding

resonance phase shift. We see that the resonance cross section has a maximum
at Eo, and the full width at half maximum of this peak is r. Eo is called the
resonance energy, and r is called the resonance width.
At the resonance energy the cross section takes the value
(JIR) = 4~ (21 + 1), (6.28)

p
which is the maximum value for the elastic partial cross section compatible
with unitarity, as seen from (XVI.2.9).
The quantity r that has emerged in these discussions can be calculated
from the phase shifts using (6.12) if these are known, e.g., from the interaction
potential in potential scattering. However, as mentioned already in several
7r:
"llt'
37r:
4
7r:
2
7r:
E
0
Figure 6.3 The Breit--Wigner cross section and its relation to the resonance phase shift.
places, the situation is more often than not the reverse: One knows the cross
section from an experiment and notices that it shows, in the neighborhood of
a certain value Eo, the features shown in the lower part of Figure 6.1. Then
one tries to fit this experimental cross section with a formula
O"i'x p = 4n(21 + 1)1 Tf(E) + Tbg(EW, (6.29)
where TfCE) is the function (6.26) and Tbg(E) is a slowly varying background
term. In this way one determines the values (Eo, 0.
Equation (6.27) is the celebrated Breit - Wigner formula; (6.26) is called the
Breit - Wigner amplitude. It hardly ever gives a completely accurate descrip-
tion of the experimental situation, which indeed one would not expect in
view of the idealizations involved. It almost never fails to give a useful
parametrization whenever resonance phenomena are involved. It is, perhaps,
the most frequently used formula of quantum physics. Its fundamental
significance comes from its introduction of the parameter r: On the level of
accuracy at which excited states are considered stationary, they are charac-
terized by one parameter,21 their energy value Eo; on the level of accuracy
at which they are considered quasistationary states, they are characterized
by two parameters, their energy and width (Eo, 0. From Chapter XII we
know that stationary states are described by eigenvectors of the energy
operator with eigenvalue Eo; in Chapter XXI we will see that, analogously,
quasistationary states are described by generalized eigenvectors of the
[essentially self-adjoint] energy operator with eigenvalue Eo + ir/2.
Like every theoretical description, the description of a quasistationary
state by a Breit - Wigner resonance (6.26) is an approximate description. If
one wants a more accurate description, one can consider r not as a parameter
but as a function of energy, r = r(E), which may depend upon one or more
parameters (e.g., the radius of the interaction and the angular momentum I).
We content ourselves here with the degree of accuracy with which the quasi-
stationary state is characterized by the two parameters (Eo, 0.
We now want to establish the connection between resonances and poles
of S/. We have already introduced, in Section XVIII.5, the name resonance
pole for a pole of SlE) in the second sheet of the energy plane immediately
below the positive real axis. As seen, e.g., from (6.23), the Breit-Wigner
formula furnishes such a pole in the lower half plane at E = Eo - ir /2.
The smaller the value of r, and consequently the sharper the resonance,
the closer this pole is to the real axis. We have already mentioned that
there is no pole on the lower half of the first sheet of the energy plane. 22
Consequently, this pole furnished by the Breit - Wigner formula must be on
the second sheet of the plane, and thus there must be a pole of Slp) in the
fourth quadrant of the p-plane close to the real axis.
21 In addition to their other quantum numbers like I, '1.

22 We did not give a complete proof of the statements in Section XVITl.5. but argued that
causality, the finiteness of the binding energies for possible bound states, and the finite range of
the interaction are sufficient, though certainly much less is necessary.
On the other hand, every simple pole of Slp) at the point P = PR =

kR - iKR in the fourth quadrant near the real axis (K R ~ k R) may, but
does not necessarily, lead to the resonance phenomenon. This is easily
seen by expanding stCp) around this pole in a Laurent series. In a sufficiently
small neighborhood of the pole, SI(P) may be approximated by the principal
part of the Laurent series ~ l/(p - PR)' However, stCp) must be unitary
[lstCp)1 = 1], so the factor multiplying 1/(P - PR) must be chosen to be
P - Pk if the rest of the series is to be expressed as an exponential ei21'1. Thus
in the neighborhood of the pole the expansion of SI(P) compatible with
unitarity is
SrCp) = ei2ill(p) ~ ei2il(p) P - Pk = ei2;I(p) P - kR + ~KR . (6.30)

P - PR P - kR - ZKR
Multiply denominator and numerator by p/m and note that
1 p2 pkR p2 k~ (p - kR)2 p2 k~ K~
pcp - kR ) - = - - - = - - - + ~ - - - +-
m m m ~ ~ ~ ~ ~ ~
(6.31 )
for real values of P that differ from kR by approximately KR ~ kR' i.e., for
values of P that fuifillip - kRI ~ K R. Then with
(6.32)
one obtains
S (E) = ei2Y1 (E) E - ER - iY/2 .h Y = _ 2K RP = Y() (6.33)

I E _ ER + iY/2 WIt m p.
With the same accuracy as the approximate equality (6.31),

2KRkR
yep) ~ - - - == r. (6.34)
m
Then (6.33) is identical with the resonance formula (6.23), and consequently
a pole of SI(P) at PR describes a resonance.
The above considerations would suggest a one-to-one correspondence
between poles of the S-matrix immediately below the positive real axis
and a resonance phenomenon characterized by (6.l); however, they do not
prove such a correspondence, because we have always assumed that the VI
is a slowly varying background phase shift. If VI varies rapidly without
having the Breit-Wigner form (6.24), then it may compensate the effect of
the pole, and the total phase shift h[ = h~R) + VI may be slowly varying. It
can, indeed, be shown 23 that phase shifts }II can be constructed that approach
thefunction (6.24) as close as one wishes in an interval Eo - .1 :5: E :5: Eo + .1,
23 L. Fonda, Fortschr. Phys. 20, 135 (1972).

XVIII.6 Resonance Scattering-Breit~Wigner Formula for Elastic Scattering 485
L\ ~ r, and that therefore describe a resonance phenomenon (6.1), but that

are not associated with a singularity of ei2Y1 • Thus choosing SI = ei2Y1 gives
an S-matrix with a resonance phenomenon but without a pole, and choosing
YI = -1'1 in (6.33) gives an S-matrix with a pole but without a resonance
phenomenon.
However, if the fourth and higher derivatives of (jl are small, then, as we
saw in the first part of this section, YI will be slowly varying. Therefore, one
generally assumes the one-to-one correspondence between poles immediately
below the real axis and resonances.
We also have seen above that a quasistationary state characterized by a
sharp maximum in the time delay is a resonance. Further, as from (6.19),
(6.24), and (2.26) it follows immediately that
1 l ( ) = d<5 l (E) = d<5\R) dyz = r/2 dyz(E) (635)

2t D E dE dE + dE (E 0 _ E)2 + r 2 /4 + dE' .
we see that a resonance leads to a sharp maximum in the time delay for the
resonating partial wave. Thus a quasistationary state, a resonance, and a
pole of Sl immediately below the real axis are in fact one and the same phe-
nomenon.
We shall now discuss the connection between the width r and the time
delay. According to the discussions in Section XVIII.2, (6.35) is the "time
delay for a monoenergetic beam;" that means the energy spread L\E of the
beam is much smaller than the width of any structure in the cross section,
L\E ~ r. If a resonance occurs, then the derivative of the background phase
shift is negligible, so that the resonant time delay is well approximated by
1 l r; _ d<5\R) _ r /2 (6.35')
2tD(~) - dE - (Eo - E)2 + r 2 /4'
The "time delay for a monoenergetic beam" that has energy E = Eo identical
to the resonance energy is therefore
4
l
tD(E)IE=Eo = r' (6.36)
This idealized time delay is not really a physical quantity, because the
monoenergetic beam is a steady state and the time delay is not an observable
quantity, as mentioned already in Section XVIII.2.
In the opposite extreme, L\E ~ r, the time delay is essentially determined
by the experimental energy distribution of the beam, as we see by inserting
(6.35) into (2.32):
2 fdE F(E ) r /2 (6.37)

l
tD -
_
- EA (Eo _ E? + r 2 /4'
Noting that for L\E ~ r one can take r/2 -> ° in the integral, and using
r/2
(Eo _ E)2 + r2/4 -> n<5(Eo - E) for r -> 0,
we obtain
(6.38)
For a Gaussian energy distribution of the incident beam,24 with !1E ~ r,
F(E _ E ) = _1__1_ e-(E-EA)2/2(~E)2 (6.39)
A .jhr, !1E '
one obtains at the resonance energy
I M- 1
tDIEA=Eo = v 2n !1E' (6.40)
And for a Lorentzian energy distribution of the incident beam,24
(6.41 )
with !1E ~ r, one obtains

(6.42)
In general, if the energy distribution in the incident beam caused by the

resolution ofthe apparatus has a width!1E of the same order as the resonance
width r, one obtains a time delay that depends on r and on the experimental
energy distribution. A particular situation arises when the experimental
energy distribution is Lorentzian [Equation (6.41)J with an energy spread
that is equal to the resonance width !1E = r. Then one obtains by inserting
(6.41) into (6.37)
tb = 2/r (= 2fi/r in conventional units of time). (6.43)
Thus we have seen that the time delay has something to do with the inverse
of the resonance width. 25 For an idealized monochromatic beam this relation
is given by (6.36). But for more realistic situations the effect of the energy
resolution ofthe incident beam will result in other connections between time
delay and width-e.g., the one given in (6.43).
So far we have considered for the parameter r, which we introduced in
(6.10), only the value r = 1. However, according to (6.22) it can take any
integer value.
The case r = 1 seems to be the only case that is realized in quantum-
scattering experiments, though some time ago there was thought to exist in
24 The normalization of F has to be chosen so that F(E - E A ) ---> o(E - E A ) as tlE ---> 0
according to (XIV.5.36). Note that F is the energy distribution of the incident noninteracting
beam at t = 0, i.e., the beam has been prepared in the remote past in such a way that an energy
measurement at t = 0 with no interaction present would give the result F(E - E A ) for the
probability distribution to measure E.
25 As we shall see in Chapter XXI, the inverse resonance width is the lifetime of a decaying
state.
XVIII.7 The Physical Effect of a Virtual State 487
Figure 6.4 Illustration of the energy dependence of the scattering cross section (6.45)
for a dipole resonance.
particle physics some evidence for the existence of a "dipole," which is a

quasistationary state with r = 2. We shall discuss this case briefly here.
With r = 2 we obtain from (6.14), ignoring the background (Yl = 0),
Sl(E) = (Eo - E + ir/2)2 (6.44)

Eo - E - ir/2
and therewith for the lth partial cross section
(r = 2) _ 4n(21 + I) r2(E - £0)2
(il (£) - p2 [(£ _ £0)2 _ r2/4]2 + r2(£ _ £0)2' (6.45)
The shape of the cross section for such a dipole resonance is depicted in
Figure 6.4.
This is to be compared with Figure 6.l(a). When background scattering is
present, the cross section will be modified accordingly.
XVIII.7 The Physical Effect of a Virtual State
Besides the bound-state poles and the resonance poles, we mentioned in

Section XVIII.5 the virtual-state poles on the negative imaginary axis of the
momentum plane. If a virtual-state pole is sufficiently close to the real axis
in the momentum plane, it can have an appreciable influence on the cross
section at low energies.
Denoting the momentum value of the virtual state pole by Pv = - iK,
K > 0, we can write the I = 0 S-matrix element in the neighborhood of the
pole as
e2iIlO(p) = So(p) = _P - p~ = _ p - ~k. (7.1)
p - Pv P + lk
This follows by arguments similar to those leading to (6.30): The principal
part of the Laurent series '" I/(p - Pv) has to be multiplied by a factor
(p - p~) such that So(p) is unitary, i.e., !So(p)! = I. The remaining phase
factor has to be chosen ( - I) so that
1 ik - p
bo(p) = -: In -.- -
21 Ik + P
fulfills the causality condition (3.15). From (7.1) follows
~
dOo(p)
= p2
k
+ k2 > 0 > -
(1
+
)R 2p ,
whereas the choice of the opposite sign in (7.1) would have lead to
doo(P) - k
~= p2 +k2
which can not always fulfill (3.15).

The Oth partial-wave amplitude is obtained by inserting (7.1) into (XVI.2.8):
-1
To(P) = -.-k' (7.2)
Ip -
The Oth partial cross section is then
2 4n
0'0 = 4nl Tol ~ -2--2' (7.3)
K +P
In the limit of zero scattering energy, when the virtual pole is close to the
real axis, this goes into
4n
0'0 ~ 2' (7.3a)
K
which is identical to the total cross section since, according to (XVII.6.6),

higher partial waves can be ignored at low energies. Thus we see that for a
virtual-state pole close to the real axis the scattering cross section can be
extremely large. Also large is the scattering length, obtained according to
(XVII.6.10) from (7.2) as:
(7.4)
K
It is easy to see that a bound state with sufficiently low binding energy
will also lead to (7.3) for the cross section. Therefore, a virtual state can only
be detected in the low-energy scattering cross section if low-energy bound
states are absent.
Virtual states have been observed in nature. In the case of proton-neutron
scattering there exists a bound state, the deuteron (isospin 0, angular momen-
tum 1). The scattering length is a~r 1) = 5.4 X 10- 13 cm. For angular
momentum equal to zero (isospin 1), no bound state is present, but ab1 ,O) =
- 23.7 X 10- 13 cm. Thus, there is a virtual state in the singlet (angular
momentum 0) p-n system, whose position is closer to the threshold than the
triple bound state and which dominates low-energy scattering.
Virtual states also cause long time delay. As we are at low energies, the
phase shift can be obtained from (cf. XVII.6.15)
p cot oo(P) = (7.5)

XVIII.8 Argand Diagrams for Elastic Resonances 489
From this we obtain

d(jo ao
dp 1+p2a6=~1+p2/1(2·
Thus the average time delay (2.26) for a virtual state with scattering length
(7.4),
t~irt) = 2 ~ ~ ---..---,- (7.6)

PI( 1 + p2/1(2'
can reach very high values. An illustration of this effect is obtained if we com-
pare the time delay with the time that a noninteracting particle moving with
a velocity p/m spends in the region of radius R, which is given by (cf. Section
XVIII.3):
T cl = m· 2R (7.5)
p
Therefore the ratio of the time delay caused by a virtual state to this classical
time delay is
t~rt ao
(7.7)
T cl = RI( 1 + p2/1(2 = - Ii -1-+-p2"a""'6 .
Depending upon the value of I( and R, this ratio can be very high. E.g., for
thep-n virtual state,takingR ~ 2.5 x 10- 13 cmandao ~ -24 x 10- 13 cm,
we see the time delay caused by a virtual state is an order of magnitude
larger than the classical time.
A virtual state is thus not simply a fictitious, mathematical state, but a
physical state in which projectile and target spend a considerable length of
time together.
XVIII.S Argand Diagrams for Elastic Resonances and

Phase-Shift Analysis 26
The Argand diagram introduced in Section XVI.3 is a useful tool for the
detection and display of resonances and the determination of their param-
eters (E R' 0. To understand this we investigate in detail the behavior of the
scattering amplitude in the Argand diagram. We first study the idealized
case of an elastic resonance without background. For this the scattering
amplitude is given by (6.26), which we write in the form
pTIR)(E) = _1_., (8.1)

E-/
26 G. Bialkowski has helped me with the writing of this section.

Im(pT,(p)]
( = -I
(E = E~ +~)
€ = -2
-4
Re(pT,(p»
Figure S.l Argand diagram for the resonant elastic scattering amplitude: p7;(P) =
(E - 0- 1 and cot bl R ) = E.
where we have introduced the abbreviation
(S.2)
Figure S.l shows the Argand diagram for Equation (S.l). As we see, this
elastic Breit- Wigner amplitude lies on the unitary circle. The top of the circle
corresponds to the resonant energy f = 0, at which value the phase shift
61(E R ) is nl 2 (cf. Figure XVI.3.1). With increasing energy E, pTf moves in a
counterclockwise sense-slowly along the lower part of the circle, and more
rapidly as E gets closer to the resonance value E R' This is indicated by the
values for f on the path of pTf in Figure S.l.
To take the background into account we use (6.25) and obtain
pI! = e i y, sin )I I + e 2 i)" pT/R) = pTjb g ) + e 2i Y'pTjR), (S.3)
where we have defined the background amplitude Tjb g ) in analogy to

(XVI.2.S) by
Tjb g ) = ~ ei" sin )II = ~ Ce i2 " - 1). (S.4)

P 21p
From (S.3) we see that the effect of the background is to rotate the resonant
amplitude by the angle 2)11' The quantity pI! for constant background ampli-
tude is shown in Figure 8.2. The resonance circle starts at the point Band
traverses the unitary circle, the only difference with the case of Figure 8.1
being that now the resonance energy is not at the point where the resonance
amplitude is pure imaginary, but at the point diametrically opposite B. If
the background changes with energy, then the point of resonance is not well
defined.
XVIII.8 Argand Diagrams for Elastic Resonances 491
Im(pT,(p)]
Re(pT,(p)]
Figure 8.2 Argand diagram for the resonant elastic scattering amplitude in the presence
of background: 1/ = T~R) + T~bg).
As we have seen above, the resonance amplitude has a characteristic

circular path in the Argand diagram, the resonance energy being the value
where the phase shift changes most rapidly. As we shall see in Section XX.3,
this feature will remain even when inelasticity is taken into account, only
there the circles become smaller and distorted.
The Argand diagram can be used to display the energy and width of a
resonance for a particular partial wave. For this purpose one has just to
draw pI; as function of energy in the Argand diagram. If pI; runs along a
circle close to the unitary circle, one knows that this particular partial wave
has an elastic resonance, and the energy value at which the amplitude has its
most rapid change will be the resonance energy ER' If there is no background,
ER will be the point at which pI; crosses the imaginary axis. In order to
draw pI;(E) one has to determine it from the experimental data, and this is
the main difficulty of phase-shift analysis. In the following we will give a
brief description of this procedure.
The data determined in an experiment are the cross sections: the total
cross section at various momenta O'(p) and the differential cross sections
dO'/dO(8, p) at various angles and momenta (or the differential cross sections
dO'/d cos 8 = 2n dO'/dO). According to (XVI.1.36) these cross sections are
connected to the partial-wave amplitudes by
dO' (p, cos 8)

d-----n = 2n 1 L
00
(21 + l)Pl(COS 8)I;(P) 12 (8.5)
cos u /;0
And by the optical theorem (XVI.2.20),

4n
L
00
O'(p) = - (21 + 1) 1m I;(p). (8.6)

p /;0
These sums extend to infinity and would be of little practical use for the
determination of the partial-wave amplitude 1/(p). Fortunately, however,
according to (XVII.6.6) the higher partial-wave amplitudes for a particular
value of the momentum p can be neglected, so that for a particular value of
p the sum in (8.5) extends in fact only over a finite number of terms:
e (p, cos e) = 211: / L (21

-d(J
d L + l)Plcos e)1/(p) /2 , (8.5')
cos /~o
and so does the sum in (8.6). How high a value of L one has to take for a
particular value of p depends upon the problem under investigation and has
to be determined empirically by the "principle of stability." One does this
by choosing some value L o . Then one determines, by the procedure described
below, To(p), ... , T Lo(P); chooses L o + 1; and determines from the same set
of data a new set of amplitudes To(P), ... , TLo(P), TLo+l(P), If 1/ and 1;
determined in this way differ significantly, L o was chosen too small for that
particular value of energy and one has to try a higher value.
Using the relation for the Legendre polynomials 27
/+I'
p/(cos e)PI'(cos e) = I <10 l' 0 ILO)2 PL(cos e), (8.7)
L~I/-/'I
where <I m l' m' 1 L, m + m') are the SO(3) Clebsch-Gordan coefficients, one
can write (8.5') in the form
d(J 2L
- e
d cos = 211: L
j~O
Picos e)c/p). (8.8)
The inversion of (8.8) is 28

21 + 1 f+ 1 d(J(p, e)
c/(p) = -4- d cos e d e Plcos e). (8.8')
11: -1 cos
For the sake of definiteness let us do this for the particular values L = 1
and L = 2. For L = 1, (8.5') becomes
d(J e
d cos = 211: {I Tol 2 + 3(To Ti + T6 T1)P 1 + 91 T112Pi} (8.9 1)
For L = 2 it becomes
+ 6 Re(T6T1 + 2TiT2)P 1(COS e)

+ [10 Re (T6T2) + 61 Td 2 + 57°1 T2 12JPicos e)
+ 18 Re (TiT2)P3(COS e) + 97°1 T212Picos e)}, (8.9 2 )
27 This is a special case of (VIU.20).

28 This is obtained by multiplying (8.8) with P,(cos IJ) and using (VIU.16) with (VII.3.25).
XVIII.9 The Effect of Background and Finite Energy Resolution 493
where the numerical factors are determined from the values of 21 + 1 and
of the Clebsch-Gordan coefficients. Thus, for example, for L = 2,
Co(P) = ITol2 + 31 TII2 + 51 T2 2, 1
etc. (8.10)
Now if da/d cos (J (p, cos (J) has been measured at a particular value of p,
then one calculates the Ct(P) from (8.S'). These "experimental" Ct(P) one
uses in equations like (8.10) to determine the" experimental" values of the
partial-wave amplitudes. There are 2L + 1 equations (S.lO). In addition
one has the optical theorem
4n
L:
L
a(p) = - (21 + 1) 1m 1l(P). (S.6')
p t=o
So one has 2L + 2 nonlinear equations for the 2L + 2 real quantItIes
1m 1l(P), Re 1l(P), I = 0, 1, ... , L, or the 2L + 2 quantities c5 t (P) and '1t(p) of
(XVI.2.10).
If the scattering at that particular energy is elastic, then one has only the
L + 1 unknowns c5/(p), and one need not consider the Cj(P) for; > L. There
are then L + 1 equations for L + 1 unknowns in (8.10). [The optical theorem
(S.6') is then not independent of (S.10).]
One has to go through the above procedure of determining the partial-
wave amplitude for every value of p that is plotted in the Argand diagram.
For small momenta, small values of L (e.g., L = 0) are sufficient, but as the
momentum increases, higher values of L have to be chosen. As the equations
are nonlinear, the solutions are not unique. Further, the experimental data
have errors. The uncertainties that result can be overcome partially by
considering p ll(p) at several values of p and requiring that it change smoothly
with momentum. Sometimes further relations, such as dispersion relations,
are used to obtain this smoothness. In this way one obtains a picture of pll as
a function of p.
XVIII.9 Comparison with the Observed Cross Section: The

Effect of Background and Finite Energy Resolution
Due to the effect of the background in the resonating partial cross section
and the background of other nonresonating partial cross sections, the
experimental cross section near resonance does not usually look as simple as
in Figure 6.1a. We tirst consider in detail the interference between the reso-
nance and the background.
If we insert (8.3) into (XVI.1.38), we obtain for the resonant partial cross
section
(9.1)
We introduce the Fano shape parameter 29
q= + cotYb (9.2)
q is called the profile index or shape profile parameter, and £ is defined as in

(8.2). After some straightforward calculation one obtains for (9.1)
(JI [1 .
= 4n(21 + I) p2 sm 2 YI + ITIR I2(q2 q2
+. 2£q -
+1
I)J . (9.3)
This is rewritten as
_
(JI -
(bg)
(JI + (JI(R) (q2 +q 2 2£q
+1
- 1) ' (9.4)
where
4n 1
(JI R ) = 2 (21 + 1) -2--1 (9.5)
P £ +
is the Breit-Wigner cross section of (6.27), and (JIb g) has been defined in
analogy to (XVI.2.9) by
(Jlbg) = 4~ (21 + 1) sin 2 YI' (9.6)

p
Inserting (9.5) and (9.6) with (9.2) back into (9.4), one can write the resonant
partial cross section also as
4n (q + £)2 4n . 2 (q + £)2
(JI = 2 (21 + 1) (2 1)( 2 1) = 2 (21 + 1) sm Yl 2 l ' (9.7)
P q+ £+ P £+
or
_ (bg) (q + £)2 (9.8)
(Jl - (Jl
£
2
+1
The factor (q + £)2/(£2 + 1), which gives the shape of the resonance, is
shown for various values of the shape profile parameter q in Figure 9.1.
We see that the shape of the resonance in the resonant partial cross section
depends considerably upon the value of the parameter q, which is connected
with the background phase shift YI' (If the background phase shift depends
upon the energy £, which is generally the case, then the shape becomes
even more complicated.) For positive values of q we see that for increasing
energies E, i.e., for increasing values of - £, the cross section increases above
background, then falls sharply at the value of the resonance energy £ = 0,
is zero at £ = -q, and increases thereafter.
For negative values of q the cross section decreases with increasing
energy, reaches zero at £ = q (i.e., before resonance), and then increases
sharply.
29 The original parameters introduced by U. Fano [Phys. Rev. 124, 1866 (1961)] were the
negative of the parameters defined by (9.2).
We are now in a position to explain the behavior of the experimental data

depicted in some of the figures of Section XVIII.l. As we have so far only
discussed the effect of the background, with the present results we can only
interpret those experiments in which the energy distribution in the beam
and the resolution of the detector can be ignored. These are the high-resolu-
tion experiments whose results are depicted in Figure 1.3 and Figure l.4(d).
These experiments measure the transmitted current, which is proportional
to the incident current minus the total cross section. As the total cross section
is the sum of the resonant lth partial cross section (9.8) plus the slowly
varying background cross section from the other partial waves, a dip in the
resonant partial cross section should show up as a peak in the transmitted
current, and a peak in the resonant partial cross section as a dip in the
transmitted current. According to the results of Figure 9.1, the resonance in
10·0 , - - - - - - - -- -- - -7"'Il": - - - - - - - - - - - - - - - ,
,
9·0 I
q = - 31
'I = +3
R·O
I
I
I
7'0 I
I
I
:;- 6·0 I
I
+ I
::
I
-: 5·0
+
1'1 =- 2 \
2" I / \
I I
4·0 / /
/ I
I /
/ /
// /
/ /
/ /
_..... ..... /
//
/'
--- -=---- -;; =--

1·0 ./
4 = - 1 ..-
...... (\.15
....... ~~ ;:. -
1.0 _.=-..::-=-
_ - -.- :- - - _ _ ..!!..·25
- ~·o
€
Figure 9.1 The resonance profile function for different values of the profile index q.
The arrow above E indicates the direction in which E increases. [From Massey &
Burhop (1969), vol. 1, with permission.]
the Ith partial cross section should show up in the following way:
For q > 0, as a dip around resonance followed by a peak right after
resonance.
For q < 0, as a peak right before resonance followed by a dip right after
resonance.
In the high-resolution experiment (very narrow spread of the electron
beam) of Figure 1.4(d), and in particular in the very high-resolution experi-
ment of Figure 1.3, one can see that the situation described for q < is
observed. (If one knew the exact value for E R , one could even infer the value
°
of q). Thus our hypothesis of Section XVIII.1, that in the collision process
(1.4) a quasistationary state He- is formed at around 19.31 eV, is justified.
We shall see that the other experiments of Figure 1.4 and Figure 1.5 lead to
the same conclusion. For these experiments the resolution of the apparatus
is coarse, and the quantity measured is not the cross section at a particular
energy value, but the total cross section in a whole energy interval.
The total cross section (XVI. 1.33) is written as
a = al + I ai' = a l + ab, (9.9)
1'*1
where a l is the resonant partial cross section given by (9.8) (describing the
resonance and the background) and it is assumed that there is no resonance
of any angular momentum l' other than I. The background cross section ab
is then given by
4n L...
ab = 2 " (If
2 + 1).
sm 2 YI" (9.10)
P 1'*1
where Yz,(E) are slowly varying background phase shifts.

Inserting (9.10) and (9.8) into (9.9), we obtain for the cross section
_ (bg) (q + f)2 + (9.11)

a - az 21 ab,
f +
where ajb g ) and ab vary slowly with energy.
At energies far away from the resonance energy, f ~ 00, we obtain from
(9.11)
(9.12)
where we call a(nr) the slowly varying nonresonant background. The maxi-
mum of the cross section (9.11) occurs at f = + l/q, and at this maximum
the cross section is
(9.13)
Because of (9.12) the cross section (9.11) is conveniently written in the

form
q2 + 2qf - 1 q2 + 2qf - 1 (9.14)

+ ab + +
(bg) (bg) (bg)
a = al 1 al = a(nr) 1 ai'
+f 2 +f2
9·0
8·0
7·0
6·0
5·0
~
+
;::,;
4·0
I
~
'"+""
JO
2:
2·0
1·0
0·0
- 1·0
E
Figure 9.2 The function that determines the form of the resonance effect relative to the
smooth background. [Massey & Burhop (1969), vol. 1, with permission.]
which shows that the resonance occurs above a gradually varying back-
ground O"(nr) and has the characteristic shape determined by the factor
(q2 + 2q£ - 1)/(1 + (2). Thus, if there is background from other partial
waves, then in distinction to (9.8), the cross section no longer vanishes at
£ = q but has its minimum value there, O"min = O"b' which lies below the
value for the background O"(nr). The shape is now determined by the shape
factor shown in Figure 9.2.
Iflql<l,
in < ')I, < :in,
then the resonance effect results in a reduction from the background on
one side of the resonance that is greater than the increase above the back-
ground on the other side. For Iq I > 1 the reverse is true.
So far in our discussions we have always assumed that the energy resolution
of the experimental apparatus is ideal. This means that the initially prepared
beam can be considered as ideally monoenergetic, so that the energy distri-

bution function can be described by the 6-function according to (XIV.5.36):
(9.15)
where EA is the central energy value in the beam. Such an ideal resolution
can be achieved in two ways: either by an ideal preparation apparatus (a
high-resolution monochromator), or by an ideal detection apparatus (a
high-resolution analyzer in the experiment of Figure 11.4.1). Both will lead
to the same result because of energy conservation. As already mentioned in
Chapter XIV, what constitutes ideal resolution depends upon the effects
that are to be observed: The energy spread of the beam I1E must be small
compared to the width of the structure in the scattering amplitude in order
that (9.15) may be used.
The observed quantity is not the cross section a(E) at a definite value E,
but the energy-averaged cross section
(9.16)
where F is determined by the energy resolution of the detection (or prepara-

tion) apparatus, because every realistic detection apparatus detects particles
not only with a definite energy EA, but with an energy in a finite interval
I1E around EA- If a(E) varies slowly as compared to F(E - E A), then
a * F :::::; a(E A), which is equivalent to using (9.15). But if I1E is not small
compared with the width r of a resonance, and consequently a(E) given by
(6.27) does not vary slowly, then one cannot approximate a * F by a(EA)'
The observed quantity is then the quantity given in terms of the T-matrix
elements by (XIV.5AO). a(E) is just an abbreviation for the expression under
the integral besides F(E - E A) on the right-hand side of(XIY.5.40) and is not
an observable quantity. Thus for a resonance of the lth partial wave with back-
ground in the lth partial wave and the other partial waves, a(E) = a(f)
given by (9.14) is not the observable physical quantity if the width r is of
the same order as the energy resolution 11E. The observable physical quantity
IS
<a(E A» = a *F = fdE a(E)F(E - E A)
= fdE (a!bg)(E) + ab(E»F(E - E A)
(9.17)
The precise value of this depends upon the energy distribution [the detector
efficiency function F(E - E A)] for the particular experiment. For the sake
of definiteness let us assume that the probability of measuring the energy E
F(E - E.. )
I
l!.E
----+------L~~----__.E
EA
Figure 9.3 Assumed energy distribution for the beam with energy resolution /t;,E ~ r.
is zero except for values in an interval of L'lE around EA' Then

I L'lE L'lE
F(E - E A) = { L'lE for EA - 2'::; E.::; EA + 2' (9.18)
o otherwise,
depicted in Figure 9.3, will describe the apparatus resolution. Equation
(9.18) is of course also an unphysical distribution like (i(E - EA ), and a
Gaussian around EA would often best reflect reality. However, for a beam
with energy spread L'lE ~ r, the assumption (9.18) is of sufficient accuracy.
With (9.18) we obtain for the observed cross section (9.17):
1 SEA + !!.E12 q2 + 2qE - 1

+ AE 1 2 O'lbg)(E) dE. (9.19)
Ll E A -!!.EI2 + E
O'lbg)(E) and O'b(E) are slowly varying functions of E. Therefore
~ O'lbg)(E A) + O'b(E A) ~ O'lbg)(E R) + O'b(E R)

is a slowly varying function of the average beam energy EA whose value is
well approximated by the value it takes when the beam energy EA is equal
to the resonance energy ER'
The second integral on the right-hand side of (9.19) is given by
«(Jr(E A )
r
= - 2A-EO'l g)(E A )
b I'A - !!.Elf q2 + 2q€ -
1+ 2
1
dE,
Ll 'A + !!.E/f €
2
EA = (ER - EA)f'
For EA far away from E R, i.e., for €A far away from zero, this will be close to
zero, as can be seen from Figure 9.2, and the averaged cross section will be
given there essentially by <anr(E A». At the value of the average beam energy
EA = ER the value of <ar(E A» is given by
r f-tJ.E/f q2 + 2qf - 1
<a/ER» = - 2f1E a\bg)(ER) !!Elf 1 + f2 df
~
2f1E
0'1
_ 00 1+f
f
~ ~ (bg) + 00 q2 + 2qf - 1 d
2 f
r
= - - a(b g )n(q2 - 1)
2f1E I .
In performing the integration we have taken the limits of the integral over
the shape factor as ± 00, which for f1E ~ r does not cause an appreciable
error, as can be seen from Figure 9.2. Inserting this and
4n I . 2
ER = (p(E R)? (2 + 1) sm (9.20)
(bg) ( )
0'1 'lI(ER)
into (9.19), we obtain for the value of the averaged cross section when the
beam energy EA is set at the resonance energy ER
<a(ER» = <anlER» + <ar(ER»
2n 2 r
= <anr(ER » + (P(ER»2 f1E (21 + 1) cos 2YlER)' (9.21 )
Thus if a value of <a(ER» is observed in a cross-section measurement with

an apparatus with poor energy resolution, then it may be above or below
or even equal to the slowly varying background <anr(E A», depending upon
the value of the background phase shift in the resonant partial wave, YI(E R ) ;:::::
I'l(E A). If cos 2Yl > 0, then the resonance will show in the observed cross
section <a(EA» as a peak above background <anr(E A» at the value EA = E R.
°
Ifcos 2Yl < 0, then the resonance will show as a depression below background.
For cos 2Yl ;::::: the resonance will not show up at all in the observed cross
section.
With these results we can interpret the experimental data in Figures lA(b),
1A(c), and 1.5, which were taken in an experiment with energy spread of the
electron beam, f1E, larger than the resonance width. Figures 1.4(b) and 1.4(c)
show an increase of the transmitted current and a decrease of the total elastic
cross section around the value EA ;::::: ER ;::::: 19.3 eV. This is what one expects
from (9.21) if the background phase shift Yl satisfies cos 2y/(E R ) < 0, or
(9.22a)
The same effect is observed in the averaged differential cross section at all
angles depicted in Figure 1.5.
From the data of the high-resolution experiments in Figure 1.3 and Figure
1.4(d), we concluded above that q < 0. Using (9.2), this means that
(9.22b)
Thus combining (9.22a) and (9.22b), we conclude as a result of our qualitative

considerations that
(9.23)
This shows that the background phase shift in the resonant partial wave for
the process e + He --+ He - --+ He + e is not at all small. Taking just Tf
of (6.26) as the Ith partial scattering amplitude and oy) of (6.27) as the cross
section around resonance energy would certainly be unjustified.
We have now explained the experimental data for the elastic scattering
process e + He --+ e + He as a resonance of narrow width around 19.3 eV
interfering with the background phase shift fulfilling (9.23). In order to
obtain more information about this He - resonance, we turn to the experiment
of Figure 1.5. This is a low-resolution experiment with dE = (50-100) x 10- 3
eV in which the intensity of the elastically scattered electrons is measured as a
function of average beam energy at various values of the scattering angle,
e = 10°, 40°, 60°, 90°, 110°. Thus Figure 1.5 gives us the averaged elastic
differential cross section as function of scattering angle and energy:
where F is the energy-resolution function of the apparatus in this experiment.

With the experimental data of Figure 1.5, one can perform the phase-shift
analysis as described in the second part of Section XVIII.S. Assuming that
in the energy range considered the partial-wave amplitudes 1/ with I > 2
can be ignored, one can use (S.9) for the determination of the phase shifts
(jo, (j1, and (j2 at various values of the momentum. As the inelasticity co-
efficients 111 are zero in the present case, where only the elastic channel is open,
the 1/(P) of (S.9) are connected with (jl(P) by (cf. XVI.2.S)
(9.24)
The Legendre polynomials that occur in (S.9) are
Po = 1, P1 = cos e, P 2 = ~(3 cos 2 e- 1),
P 3 = t(5 cos 3 e- 3 cos e), P4 = i(35 cos 4 e- 30 cos 2 e + 3).

(9.25)
One can first ignore the partial-wave amplitudes above T1 and use (S.9 1).
e
At = 90° one has, according to (S.9 1), only the contributions from To.
As according to Figure 1.5 there is still a resonance effect at e = 90°, one
concludes that the resonance occurs in the S-wave:
(9.26)
e
One can now apply (9.21) to the cross section da/d0.(E A , =7 90°). Instead
of the apparatus resolution function (9.1S), one may have to take a different
one, depending upon the experiment. 30 Then the expression on the right-hand
side of (9.21) will be different, but the principle of determining the resonant
phase shift 60 = n/2 + Yo is the same. The curve in Figure 1.5 for () = 90°
respresents
It shows again the decline below the nonresonant differential cross section
characteristic of a resonating partial wave with background phase shift
given by (9.22a). From the values of
(:~ (EA' () = 900»)

at various energies EA around the resonance energy ER , one can determine
the background phase shift Yo and the true resonance width r. The values
obtained from this experiment are
r ~ (15-20) x 10- 3 eV, (9.27)
Yo is in agreement with the restrictions (9.23) obtained from the other experi-
ments above.
One may now use the data of Figure 1.5 at () = 60°. Picos 60°) is close
to zero, and the influence of a possible T2 term, which would be considerable
for P2(COS () i= 0 according to (8.9 2 ), is minimized. Fitting (8.9 1 ) to the
experimental data for () = 60° with 6~ = n/2 and Yo = 100°, one obtains
(9.28)
To check that the contribution from the higher partial waves is small, one
can now fit the experimental data for the other scattering angles () with
(8.9 2 ), For example, with Yo = 100° and Y1 = 25°, one obtains from the
experimental curve for () = 10°
(9.29)
This is indeed small, and justifies the working hypothesis made at the begin-
ning of our analysis that T2 and all higher partial waves can be ignored.
The He in the collision process
e + He ---> He- ---> He +e (1.4)
is in the ground state lSo with I = 0, s = O. The e in this process has s = t,
and as we know from (9.26), I = O. Consequently, the quasistationary state
He - has I = 0 and s = 1, i.e., it is a 2S 1/2 state. It is very reminiscent of the
30 Most frequently the apparatus resolution function is a Gaussian (6.39). The observed
cross section is then according to (9.17) the convolution of a Gaussian and a Lorentzian func-
tion, which is known as the Voigt integral or Voigt profile. The Voigt integral cannot be found
analytically, and has been computed and tabulated. The Lorentzian width r and the Gaussian
width !'J.E can be obtained from a fit of the experimental data with the Voigt profile. See, e.g.,
H. G. Kuhn, Atomic Spectra, Academic Press, 1969, Section VIID.
Problems 503
2S i / 2 ground state in the lithium spectrum (cf. Figure VII.2.1). The only
difference is that the charge of the nucleus in lithium is + 3, whereas that in
He - is + 2, and the He - state has a width different from zero and given by
(9.27).
Problems
1. For the infinite repulsive square-well potential of Problem XVII.3, obtain the time
which the projectile spends in the interaction region !Jl of radius a and the time
T(91)1 = 0
delay tb=o caused by this potential. Discuss and interpret this result.
2. (a) Obtain the S-wave phase shift and the 1= 0 S-matrix element So(P) for the
square-well potential.
(b) Find the energy values of the bound states of angular momentum 0 as poles of
So(p). Discuss the relations for the mass, the depth Vo , and the radius R of the
well that have to be fulfilled in order for there to be (i) no bound state, (ii) one
bound state, (iii) two bound states.
(c) What are the virtual states of angular momentum zero? Under which conditions
do there exist no virtual states for I = O?
(d) Consider a very deep well fulfilling 2mVoR2 ~ I. Show that there are regularly
spaced s-wave resonances. Determine the energy and width of these resonances.
(e) Obtain the scattering matrix element So(p) at energies close to a resonance
energy, and obtain the background phase shifts. Calculate the elastic scattering
cross section (I = 0), and show that it can be written as the sum of three terms:
one from background scattering, one from resonance scattering, and one from
the interference between resonance and background. Give a graphical
representation of the elastic scattering cross section as a function of energy.
3. Calculate the time delay caused by a narrow resonance of angular momentum
IR n
and with resonance parameters (E R , of a beam with an energy distribution given by
. I
L <E 1131 W"'IE 11 3 ) = lV;(E) = --- F(E - E R ) for every I,
h peE)
where
otherwise,
with I1E ~ r. Give the result in the approximation of constant background phase
shifts.
4. Calculate the time delay of a beam with a Lorentzian energy distribution (6.41)
caused by a resonance with resonance parameters (Eo = E A , r = I1E).
5. Taking for the scattering amplitude an expression of the form
_ I r/2
7;=-----
p (E - Eo) - i(r/2)
will also lead to a Breit-Wigner cross section (6.27). Why can f; not describe the partial-
wave amplitude for resonance scattering?
6. (a) Describe the behavior of the resonance phase shift for a "dipole" resonance
(6.44).
(b) Discuss the cross section near resonance energy as a function of energy for a
dipole resonance, taking the effect of a slowly varying background y/E) into
account. Draw graphs in analogy to Figure 6.1.
7. Compare the case of a dipole resonance with the case of two resonances close
to each other that are described by a scattering amplitude
T(p)
1
= _ ei2yr(E)
(11' 2 R
II' e (yi'(E)- yr(E))))
i2
+ -=2_X,,--~~~~_
I p E R -E-i(r R /2) Ex- E - i(rx/2)
where YR, r R, ;'x, r x are real constants: Ei ~ C > Yi (i = R, X) and IER - Ex I are of the
same orders as r Rand r x; and yf(E), y~(E) are slowly varying functions of E. Consider
first the special case of zero background shifts yf = y~ = O.
8. (a)
For a resonance, p71(E) moves counterclockwise in the Argand diagram.
Show that this is related to causality.
(b) Discuss when p71(E) can move clockwise in the Argand diagram.
9. Consider the quasistationary state with energy Eo and width r for which the
parameter r in
SI(E) = exp (i2rarctan~)

Eo - E
has the value r = 3. Obtain the slope of the resonant cross section for such a "tripole"
resonance. (Ignore the background.)
10. Explain, using the shape of the function al(E) in Figure 6.1 for a background
phase shift 1'1 = n/4, why the cross section (a(EA}) observed by an apparatus with
poor energy resolution tiE ~ r need not show a structure at the energy value EA equal
to the resonance energy E R'
11. Show that the effect of a bound state with low binding energy in elastic scattering
is the same as that of a virtual-state pole close to the real axis.
CHAPTER XIX
Time Reversal
For the discussion of multichannel resonances in Chapter XX, we require

the properties of the S-matrix that follow from time-reversal invariance.
In Section XIX.1 space-inversion invariance is applied to determine some
properties of the S-matrix. Time reversal is introduced in Section XIX.2.
The main purpose of Section XIX.1 is to introduce Section XIX.3, where
the consequences of time-reversal in variance for the properties of the S-
matrix are examined.
XIX.1 Space-Inversion Invariance and the Properties of the

S-Matrix
In Section V.4 we introduced the observable parity, represented in the space

of physical states by the unitary operator Up fulfilling the relations (V.4.l)-
(V.4.3) with the position, momentum, and angular-momentum observables.
We mentioned in Section V.4 that not all energy operators commute with
the parity operator-in particular, the interaction Hamiltonian that causes
the weak transitions, and consequently the transition operator for the weak
interaction, do not commute with Up. The weak interaction violates parity
in variance. However, many scattering and decaying systems are parity-
invariant, i.e., in addition to
(1.1)
505
506 XIX Time Reversal
they also fulfill

[V, Up] = 0 or [T, Up] = 0, (1.2)
or in terms of the S-operator
S = U~SUp. (1.3)
Parity invariance leads to some constraints on the T-matrix and S-matrix
elements. For example, it follows from (V.4.2) and (V.4.S) that
(1.4)
where n~ is + 1 or - 1, and for the angular-momentum eigenvectors (XVI. 1.3),
in analogy to (V.4.39),
U p 1E1l 3 1'J> = (-I)'n~IEIl31'J1t>. (1.5)
The additional, internal quantum numbers '1 usually do not change under
Up; thus '1" = '1. The internal parity n~ may be an element of the set of
additional quantum numbers '1.
From (1.4) it follows for the S- and T-matrix elements in the momentum
basis, using (1.3) or (1.2), that
<pll'J'ISlpl'J> = <pll'J'IU~SUplpl'J>
= <-pll'J'ISI-PI'J>n~,n~, (1.6)
or
<pi '1' I Tip '1 > = <- pi '1' IT I - P'1 >n~, n~. (1.7)
From (1.5) it follows for the T -matrix elements in the angular-momentum
basis (assuming T is spherically symmetric) that
<EI'I~I'J'ITIE1l31'J> = <EI'I~I'J'ITIE1l31'J>(-1)'+l'n~,n~. (1.S)
The conditions (1.7) and (1.S) constitute restrictions on the properties of the
T-matrix elements.
We shall also restrict ourselves to the case discussed in Chapter XVI,
namely, where the internal quantum numbers '1 do not contain internal
angular momenta. Then, using (XVI.1.lS), one obtains from (1.S)
<E I' l~ '1'1 TIE 113'1> = n~,n~<E ll~ '1'1 TIE 1131'J>b ll , b'31J (1.9)
or for the partial-wave amplitude defined by (XVI. 1.34), (XVI. 1.23),
(XVI. I .IS),
(1.10)
This shows that in the case of elastic scattering of spin-zero particles, parity
in variance does not impose any additional restrictions that are not already
consequences of rotational invariance. For '1 -=f. '1 ', (1.10) states that the
partial-wave reaction amplitude will be different from zero only if the in-
ternal parity of the initial state is equal to the internal parity of the final
state n~ = n~, or if the internal parity is conserved. Parity invariance (1.2)
XIX. 2 Time reversal 507
leads to parity conservation (1.7), (1.8), (1.10), in the same way as rotational
invariance (XVl.l.1) led to angular-momentum conservation (XVI.1.18).
XIX.2 Time Reversal
The main purpose of the preceding discussion on parity invariance is to

serve as an introduction to another kind of invariance, which superficially
appears to be very similar to parity in variance but is of a completely different
kind: the invariance under time reversal. Naively, time reversal is the re-
placement of t by t R = - t; however, this is not a physical operation like
taking the mirror image Xi ~ xf = - Xi' as we cannot turn time backwards.
But we can compare the state of a physical system having a position Xi and a
velocity Vi with the state of the physical system having the same position Xi
and the velocity - Vi' If in a scattering process all positions and velocities
undergo this transition at all times:
f7: x ~ +x, f7: y ~ -y or f7: x ~ x, f7: p ~ -p, (2.1)
then one obtains a process which looks like the original process going
backward in time.
The time-reversal operation f7 is the velocity reversal (2.1). For a quantum
physical system this transformation f7 is represented by an operator in the
space of physical states. We call this operator the time-reversal operator
AT' If AT represents a usual observable, like those we have considered so
far, then AT should be a linear operator. We shall see instantly that AT
cannot be a linear operator.
In analogy to (VA.l) and (V.4.2) for the parity operator Up, one would
translate (2.1) for the quantum physical observables Qi and Pi into
ATQiAi 1 = Qi' (2.2)
A T P i A i 1 = -Pi' (2.3)
We shall justify this below.
For the orbital-angular-momentum operator Li = Eijk Qj Pk and thence
for any angular-momentum operator J i , it follows from (2.2) and (2.3) that
ATJiAi1 = -J i . (2.4)
Let us now consider the commutation relation between the momentum
operator Pi and the angular-momentum operator J i :
[J i , P k] = iEik1P 1, (2.5)
and insert (2.3) and (2.4) into (2.5):
AT[J i , PkJAi 1 = - iEikiATP1Ai 1. (2.6)
If AT were an ordinary linear operator, then we would obtain by multiplying
(2.6) with Ai 1 from the left and AT from the right
[J i , P k] = -iEik1P 1·
This would contradict (2.5). Thus if AT is to reverse the direction of the

momenta (2.6), keeping the positions unchanged, then AT cannot be a linear
operator.
[A one-to-one mapping A of a linear space :Yf over the complex

numbers IC is called a semilinear operator if
A(ljJ + ¢) = AljJ + A¢ for all ljJ, ¢ E.Yf, A¢ = °¢ => = 0,
(2.7a)
and, for all a E IC
Aa¢ = a'A¢, where a' = a or a' = a*. (2.7b)
A semilinear operator is called antilinear if a' = a* (complex con-
jugate of a), i.e., if
Aa¢ = a*A¢. (2.7c)
A semilinear operator is called linear if a' = a. Thus in distinction
to a linear operator, already defined by (1.3.1) and (1.3.2), an anti-
linear operator also transforms the complex numbers into their
complex conjugates:
Ai = -iA. (2.7')
In a scalar-product space :Yf one can uniquely define for every
antilinear operator A (defined everywhere in :Yf or in a dense
subspace thereof) an antiadjoint operator At (often simply called
the adjoint operator) by
(Aj, g) = (Atg, f) = (f, Atg)*, j, g E JIt'. (2.8)
The following relations hold for a linear operator Land antilinear
operators AI' A z :
(AILAz)t = AiLtAt·
(2.9)
An anti linear operator is called anti unitary if it preserves the norm,
i.e., if
(2.10)
The product of two anti unitary operators is a unitary operator.
For anti linear operators one has, in analogy to the case for
linear operators,
Tr(A W) = Tr(W A), Tr(QAWAt) = Tr(AtQAW), (2.11)
where Wand Q are Hermitian operators. It is always assumed that

the operators are such that all traces that occur in the proof are
well defined.]
If we choose for the representative of the velocity reversal or time reversal,

AT' not a unitary operator as we did for the space inversion in Section V.4,
but an anti unitary operator, then (2.6) will not lead to a contradiction, but-
using (2.7c)-back to (2.5).
We therefore make the following statement: The time-reversed or velocity-
reversed state WT(t) of the state Wet) is obtained from Wet) by
WT(t) = AT W(t)AL (2.12)
where the time-reversal operator AT is an anti unitary operator
A~AT = I (2.13)
that fulfills the defining relations (2.2)-(2.4).1
In addition AT fulfills
A~ = £I, where (. = + 1 or - 1. (2.14)
Furthermore we require that the energy and transition operators commute
with the time-reversal operator AT:2
AT T - TAT = 0, (2.15)
and that the S-operator fulfills
(2.15')
(2.15) and (2.15') are the statement of time reversal invariance.
The justification of this statement consists of two parts. The first and more
difficult part is to show that the operation of time reversal is represented
by a semi unitary operator [i.e., an operator fulfilling (2.7a), (2.7b), (2.13)].
In our heuristic arguments above, we have simply assumed that-in resem-
blance to the basic assumption I-time reversal is represented by something
like a linear operator. The justification of this is, essentially, the celebrated
Wigner theorem which applies to a much wider class of transformations than
that of time reversal. This part is briefly discussed in the appendix to this
section.
The second part in the justification of the above statement is to establish
the particular properties of the time reversal operator expressed by the
relations (2.2), (2.3), (2.4), (2.15).
From the physical interpretation of the time-reversal or velocity reversal
operation, the following requirement results: If the expectation value of the
momentum in the state Wet) is 15, then the expectation value of the momen-
tum in the state WT(t) should be -15. This means
Tr(P; Wet)) = - Tr(P; WT(t)) = - Tr(P;AT W(t)AD = - Tr(Ai-P;A T Wet)).
(2.16)
1 If there exist additional independent observables that cannot be expressed in terms of

Pj' Qj, Jjo then their relation to AT also has to be specified.
2 So far, this requirement has been found to hold except for decaying KO mesons, and there
may be a superweak part of the Hamiltonian that does not fulfill (2.15).
As this is to be fulfilled for any state Wet) and its corresponding time-reversal
state WT(t), it follows that
Pi = -A~PiAT'
from which (2.3) is obtained. Equation (2.2) follows in the same way. Equation
(2.4) follows from (2.3) and (2.2), as mentioned above. It has also already
been mentioned that from (2.6) the antilinearity of AT follows.
Executing the operation of time reversal two times on all states,
(2.17)
should lead to the original state. From this requirement it follows that
ATAT = tJ with a* = 1. (2.18)
But as ATA} = ATATAT = A}A T and consequently AT I': = I':A T , it follows
from the anti linearity that I': must be real. From this (2.14) follows.
Whether one has to choose for the time-reversal operator an AT fulfilling
(2.14) with I': = + 1 or with I': = -1 has-at least at the present time-to be
decided by experience. The following choice has proved to be without con-
tradiction to experimental observations:
I': = (_l?S, (2.19)
where s is the spin of the physical system. 3
The requirement (2.15) is empirical in nature: No interaction has been
found (except for the above-mentioned K O system, which we shall ignore
here) that violates time reversal invariance as stated by (2.15). Also, if the
Heisenberg equation of motion (XII. 1.24) with (XII.1.26) is to be fulfilled
by the position operator, then it follows from (2.2) and (2.3) and the anti-
linearity of AT that H must commute with AT'
Equation (2.15') is a consequence of (2.15) and the antiunitarity of AT'
The state Wand its AT-transformed state W T may be considered as
different states of the same physical system. It is, however, more practical
to restrict the scope of the physical system and to consider all the AT-trans-
formed states of all the states of a physical system as belonging to a new
system, the AT-transformed system of the original system. To obtain an
idea about the property of the AT-transformed system, let us consider a
particular state:
WT(t) = AT W(t)A~.
From the time-development axiom V -in particular, from (XII.1.15) and
(XII.1.16)-it follows that
WT(t) = ATe-itHWoeitH A~ = e+ itH AT WoA~e-itH
= e- i(-t)HW6'e+ i(-t)H. (2.20)
3 E. P. Wigner, in GrollP Theoretical Concepts and Methods in Elementary Particle Physics

(Ed. F. Gursey), Gordon & Breach, New York, 1964, p. 37.
Thus the AT-transformed system develops forward in time in the same way
as the original system would develop backward: W( -t) = e-i(-t)HWoei(-tlH.
If an original scattering system develops from a particular in-state Win into
a particular out-state wout, then the time-reversed system develops from an
in-state (wT)in agreeing with the out-state of the original system (wTyn =
w out into the out-state (WT)out = Win.
Appendix to Section XIX.2

In this Appendix we give a brief description of the conditions and arguments
that lead to the statement that the transformation of a physical system is
described by a semiunitary operator. We choose here the transformation of
time reversal.
Let WT(t) denote the state that is obtained from the state Wet) by the
operation of time reversal :!I:
:!I: Wet) --+ WT(t).
It is natural to require that:

1. If two states have no property in common, then the :!I -transform of
these two states should also have no property in common.
2. If one state has a more specific property than a second state, then its
:!I-transform should also have a more specific property than the
:!I -transform of the second state.
Let us restrict ourselves to states that are described by projection op-
erators A1 and A z . Then requirement 1 means:
A 1A z = ° => AiAI = 0,
or
A 1£, .1 A z £' => Ai£' .1 AI£'
(orthogonal properties go into orthogonal properties). Requirement 2
means:
A 1A z = A1 => AiAI = Ai-
A1 is the state with the more specific property than A z . (As an example, A z
may be the state that has angular momentum j = 2, and A1 the state that
has angular momentumj = 2 and angular-momentum componentj3 = + 2.)
It is further natural to require that:
3. Time reversal is a one-to-one mapping in the set of states.
From these three requirements it follows by a general mathematical
theorem 4 that time reversal:!l is described by a semi linear operator AT:
:!I: Wet) --+ WT(t) = AT W(t)A}.
4 E. Artin, Geometric Algebra, Academic Press, New York, 1951, p. 88.

It is further natural to require that:

4. Overall probability should be preserved by the operation of time
reversal. This means that for all Wand their corresponding W T
Tr W= Tr W T .
From this follows (2.13).
Thus, the operation of time reversal is represented by a semi unitary
(unitary or antiunitary) operator.
So far no specific property of the time-reversal operation has been used,
and the above arguments apply to any transformation of the physical states
with the properties 1, 2, 3, and 4. Thus every such transformation is repre-
sented by a semi unitary operator which is in general determined up to
a phase factor. The celebrated Wigner theorem, 5 which states that a sym-
metry operation of a quantum physical system is induced by a unitary or
antiunitary operator, is contained in the above statement.
XIX.3 Time-Reversal Invariance and the Properties of the

S-Matrix
Before we can apply the assumption of time-reversal invariance to find the
properties of the S-matrix and T-matrix elements, we have to determine
the action of AT on the basis vectors. This is done in the same way as it was
done for the parity operator Up in Section V.4. For the momentum eigen-
vector Ip 17 >it follows from (2.14) that
(3.1)
»
Consequently, (AT Ip 17 is also an eigenvector of Pi but with eigenvalue
- Pi' If there are no additional eigenvectors and p itself labels the basis
system completely, then it follows from (3.1) that
ATlp> = a(p)l-p>,
where a is a phase factor, which, because of
A}lp> = Elp> = a(p)ATI-p> = a(p)a( -p)lp>,
a2 = E (= + 1 for the spinless case) (3.2)

and can be shown to be independent of p.
For the angular-momentum eigenvectors IE //3> it follows in the same
way, from (2.4) and (2.14), that
ATIE //3> = alE / - 13>' (3.3)
5 Proofs of the Wigner theorem can be found in Ludwig, (1954), p. 10 I; U. Uhlhorn, Ark.
Fys. 23, 307 (1963); L. O'Raifeartaigh, G. Rasche, Ann. ofPhys. 25, ISS (1963); V. Bargmann,
Ann. of Math. 59, I (1954). See also J. M. Jauch, in Group Theory and Its Applications (Ed. E. M.
Loebl), Academic Press, New York, 1968, p. 131, where further references are given.
XIX.3 Time-Reversal Invariance and the Properties of the .'i-Matrix 513
If there are additional quantum numbers Y/ (internal quantum numbers,

channel labels), then the action of AT on the basis vectors I··· Y/) depends
upon the relation between AT and y/0P. Let us assume that y/0P does not con-
tain any internal angular momenta. Let us further assume that all y/0P com-
mute with AT:
(3.4)
This will be the case if y/0P stands for the internal-energy operator of the
target, for instance. Then
where r:t = r:t(y/) may depend upon Y/ and perhaps also upon l.
Let us now consider' the S-matrix element in the angular-momentum
basis (XVI.1.42) and the reduced S-matrix elements <Y/IIS/(E)IIY/') defined
by (XVI.1.46) and (XVI.1.44). One can either use (2.14) for the T-operator
together with (3.S) and the definition of the S-matrix given by (XVI.1.41),
or one can use (2.1S) for the S-operator directly together with (3.1S), to
calculate, using (2.8) and (3.S),
<E 113 Y/IAtSATIE' l' I~ y/') = (ATIE 113 Y/), SATIE' l' I~ Y/'»*
= r:t(y/)r:t*(y/')<E 1 - 13 Y/ 1 S 1 E' l' - I~ r!,)* (3.6)
On the other hand, we obtain from (2.1S):
<E 113 Y/IAtSATIE' l' I~ y/') = <E 113 Y/IStlE' l' I~ y/')
= (S 1 E 113 Y/), 1 E' l' I~ y/'»
= <E'I'/~I'/'ISIE1l31'/)*. (3.7)
Thus
or, taking the complex conjugate of (3.8),
In order to determine the phase factors, we calculate

<E 113 Y/IAtATIE' l' I~ 1'/') = r:t(I'/)r:t*(I'/')<E 113 Y/IE' l' I~ 1'/')*. (3.10)
On the other hand, because of (2.13),
<E 1131'/IA}ATIE' l' I~ 1'/') = <E 11 3 1'/IE' l' I~ 1'/')

= p-l(E)c5(E - E')c511,c5/3/3c5~~, (3.11)
if the generalized eigenvectors are "normalized" in the usual way with real
pee). Consequently, comparison of (3.10) and (3.11) shows that we require
(3.12)
Equations (3.8) and (3.9) with (3.12) are called the reciprocity theorem 6 (for
the spinless case). This relation is quite distinct from the corresponding
relations-e.g., (1.7) or (1.8)-following from parity invariance, as it relates
two different processes with each other. The left-hand side of (3.9) describes
the process I] ----> 1]', whereas the right-hand side describes the process 1]' ----> 1].
The reciprocity theorem then states that these two processes have the same
transition probability. The measurement of these transition probabilities
provides a possibility of checking whether time-reversal in variance is
fulfilled.
We now want to determine the consequences of time-reversal invariance
upon the lth partial reduced S-matrix element <1]'IIS1(E)III]) defined by
(XVI.1.4S) and (XVI.1.43). Inserting these into (3.8), we obtain
(3.13 )
inserting them into (3.9) gives
(3.14)
Thus the lth partial reduced S-matrix element <I]'IIS1(E)III]) is not only a
unitary [cf. (XVI.2.4)] but also a symmetric matrix. We shall make use of
this fact in the following section.
For an experimental check of the reciprocity theorem we obtain the
consequences of (3.14) for the partial cross section. Inserting (3.14) into
(XVI.1.48), we obtain for the partial-wave amplitudes (XVI. 1.34) the follow-
ing relation:
(3.15)
Inserting this into (XVI.1.37), one obtains the following relation between
the partial cross sections af-q and a7- q' describing the processes I] ----> 1]' and
1]' ----> 1], respectively:
(3,16)
An analogous relation holds for the total and differential cross sections.
Equation (3.16) has been derived for the case of spinless particles. If pro-
jectile and target for the system I] have the spins sand ST' and for the system
1]' the spins s' and s~, then (3.16) is replaced by
aq- q' = aq'-q L (2s + 1)(2s T + 1), (3.16')

p,2 (2s' + 1)(2s~ + 1)
For a check of the principle of detailed balance (3.16), we consider the
experimental data of Figure 3,1.
6 The reciprocity theorem is also called the principle of detailed balance or the principle of
microreversability.
XIX.3 Time-Reversal Invariance and the Properties of the S-Matrix 515
200
A
I
I I
50 : AI"(p,p)AI"
I
20
15 B
10
5
1.1 1.2 1.7 1.8

C
3
2 Mg"(a,p)AI"
70 o
~60
+
.
~ 50 .. "
~:f -r·
I
20t2.7 ' 2 .8 3.1
,I
3.2
I
3.3 3.4
2.9
Incident particle energy 10 MeV (center of mass system)
Figure 3.1. Relative cross section as function of energy for various scattering
processes: A, elastic scattering of protons on aluminum. 8, reaction Al27 + p --> a +
Mg24. C, reaction Mg24 + a --> p + AI27 . D, elastic scattering of a-particles on
magnesium. [From S. G. Kaufmann et at., Phys. Rev. 88, 673 (1952), with permission.]
This figure depicts the counting rate (which is proportional to the cross
section) as a function of the energy for various processes. Parts Band C
are for the processes
AF? + p ...... r:x + Mg24,
Mg24 + r:x ...... p + AI 2?,

which are related by the principle of detailed balance, and (3.16) should
apply, with '1 characterizing the A12? -proton system and '1' characterizing
the Mg 24 _r:x system. The energy scale has been appropriately adjusted, and
the curves should agree with each other (except for a factor), The data were
taken from two different experiments with different resolution; still, the
agreement is convincing. The various bumps correspond to resonances of
the Si 28 system, and we shall discuss this, as well as the comparison with the
curves in parts A and D, in the following chapter.
Problems
1. Show that the product of two anti linear operators is a linear operator and that the
product of an anti linear operator and a linear operator is antilinear.
2. Show that as a consequence of the antiunitarity the time reversal operator must
fulfill A~ = d, where f = + 1 or f = - 1.
3. Show that from the definition of the M0'ller wave operators (XV.3.7 ±) it follows
that n± = A~n+ AT if Hand K commute with AT' Use this to show that (2.15) is a
consequence of (2.14) if the S-operator is defined in terms of the energy operator.
4. Prove that the time-reversal operator transforms the exact energy eigenvectors
IPI) +) into the exact energy eigenvectors I - pI)' -). (The label + or - refers to the
definition according to the Lippmann-Schwinger equation.)
5. Show that the time-reversal operator transforms an out-state into an in-state and
vice versa.
6. Consider a multichannel system of spin less particles. Show that the transformation
property under Up and AT implies that the S-matrix in the momentum basis is symmetric.
7. Show that from the Heisenberg equation of motion (XII.1.24) for the position
operator, A = Qi' with (XII. 1.26) and the anti linearity of AT follows that H commutes
with AT'
CHAPTER XX
Resonances in Multichannel
Systems
In the present chapter we discuss resonances in scattering processes from an

initial state into several possible final states, where the internal quantum
numbers of the resonances are, in general, different from the quantum numbers
of the initial state and the final states. In Section Xx. 2 we first discuss, in detail,
the case of a single multichannel resonance and then the case of a double
multichannel resonance (which occurs if there are two resonances with
different internal quantum numbers in the same partial wave, with both the
resonances coupling to initial and final states). In Section XX.3 the Argand
diagrams for inelastic resonances are described.
XX.1 Introduction
In the discussion in Chapter XVIII we assumed that there is only one channel
open in the scattering experiment and that it is this particular channel to
which the resonance phenomenon is restricted. Then there is only one
reduced S-matrix element <'1IIS,(E)II'1> that is different from zero, and it is
this reduced matrix element that resonates. This is the case of elastic scattering.
a+ T--+R--+a+ T. (1.1)
'1 are the quantum numbers of the initial state (a, T), of the final state (a, T),
and of the resonance R.
This is, of course, a very special situation, which in general will not be
fulfilled. When the energy is high enough, so that several final channels are
517
518 XX Resonances in Multichannel Systems
open, a resonance can occur not only in the elastic scattering process
described by SlE) = <IJAIISlE)IIIJA) but also in the inelastic processes
described by <IJIISI(E)IIIJA) with IJ of- IJk In order that one particular reso-
nance shows up in several channels this resonance state must have nonzero
transition matrix elements to all initial and final states with quantum
numbers IJ, IJ', IJ". If one also wants to assign internal quantum numbers, say
K, to the resonance state, then these must be different than the quantum
numbers IJ; the corresponding observables KOP and lJ op cannot commute.
It may be a matter of taste and convenience whether one characterizes
these resonance states by internal quantum numbers K = (Kl K2 ... Kn) and
considers the resonance energy as a derived quantity ER = E(K R ), KR =
(KRI KR2 ... K Rn ) determined by the internal quantum numbers, or whether
one characterizes the resonance state directly by the resonance energy E R'
The former view is prevailing in particle physics, the latter in nuclear physics.
We shall follow here the former view, assuming that the resonance state is
characterized in addition to the angular momentum by a set of "internal"
quantum numbers K, because this allows a very simple derivation of multi-
channel resonance formulas from the results of Chapter XVIII.
XX.2 Single and Double Resonances
The reduced matrix elements of the S-matrix <I]IISlE)III]') form a unitary

matrix, according to (XVI.2.3). According to a well-known theorem of
linear algebra,l for any unitary matrix there exists another unitary matrix
that diagonalizes it, and the nonzero elements of the diagonal matrix (the
eigenvalues) are of absolute value one. Thus for any value of E, there exists
a unitary matrix <Kill]) (with an adjoint <I]IIK) = <KIII]»),
<KIK') =6 KK , = I <KIII])<I]IIK'), 6~,~ = <1]'111]) = I <IJ'IIK) <Kill])
K
(2.1)
such that
(2.2)
6\")(E) are called the eigenphase-shifts or eigenphases. The matrix <Kill])

may be different for different values of E and, therefore, in general, a function
of the scattering energy. From the time-reversal invariance it follows that
the reduced S-matrix is symmetric [Equation (XIX.3.14)J. From the general
theorems 1 of linear algebra it follows that for a symmetric unitary matrix
there exists an orthogonal matrix that diagonalizes it. Thus, the <KIIIJ)
fulfill not only (2.1) but also
(2.1 a)
1 See, e,g., A. Lichnerowicz (1967) or V. I. Smirnov (1964), Vol. III, Part J, Section 42.
Inverting (2.2) with the help of (2.1), we obtain

<'1115 1(E)llrl') = L ei2oiK)(E)<'1IIK) <KII'1')· (2.3)
K
To understand the physics behind this mathematical transformation of

the reduced 5 matrix into diagonal form, let us consider the basis vectors
that were used for the space of the initial states (a, T):
IE 113 '1) = IE 113) @ 1'1)'
The orthogonal matrix <'1IIK) transforms this basis system into a new basis
system
(2.4)
with the quantum numbers E 113 unaffected but with new internal quantum
numbers K. The observables whose eigenvalues are K do not commute with
the observables whose eigenvalues are '1:
(2.5)
The quantum numbers '1 were chosen so that the initial state (a, T) is charac-
terized by a definite value of '1: '1 = '1A = '1(a, T). Now it may be that the
quasistationary state R that is formed in the process a + T ~ R ~ a + T
does not have a definite value of the quantum numbers '1 but a definite value
of the quantum numbers K, say K = KR' Then we can repeat the same con-
siderations as in Sections XVIII.2 and XVIII.6-this time not for the in-
ternal quantum numbers '1, but for the internal quantum numbers K, in
particular for that value KR of the quantum numbers K that constitutes the
internal quantum numbers of R. The time delay t~R) and phase shift <5I KR ) are
then characterized by this quantum number in the same way as in Chapter
XVIII they are characterized by the quantum numbers '1A of the initial (and
final) projectile-target system (a, T). Thus it is not for the quantum numbers
'1A of the initial state (a, T) that (XVIII.6.23) holds, but for the quantum
numbers KR of the resonance state R, which now has quantum numbers that
are incompatible with the quantum numbers 'lA' i.e., they fulfill (2.5). Thus
instead of 51 = <'1115lE)II'1A)' we now write Equation (XVIII.6.23) for
<KR 115/(E) IIKR):
<K RI15 1(E)IIKR) = ei2of'(KR)ei2y,(KR) = e i20 ,(KR)
= (1 + ER - irR
E - irR/2
)ei2Y,(KR). (2.6)
<5f(KR) here denotes the resonant phase shift, and Yl(K R) the background
phase shift, so that the phase shift is again written
<5lKR) = <5f(KR) + Y/(KR)·

The reduced 5-matrix elements <'1115 1(E)II'1') between the (reduced) basis
vectors 11'1) have then to be expressed in terms of the reduced 5-matrix
elements <KIISlE)IIK') between the (reduced) basis vectors 11K), using the
transformation
(2.4')
IC
which leads to (2.3) with

(\K·ei2b\"') = <KIIS,IIK').
The matrix element <Kill]) is the number that characterizes the strength
of the coupling between the states with the internal quantum numbers I]
and K. As mentioned above, they depend in general upon the energy E.
It may of course be that a quasistationary state is formed not only for
one value KR of the quantum numbers K, but also for other values K =
K Xl ' K X2 "" • This may even happen at the same value I of the angular mo-
mentum and even at close energies ER and E X, . We assume for the sake of
definiteness that there are two values KR and Kx for which quasistationary
states are formed with angular momentum I, and their energies and widths
are (ER' lR) and (Ex, Ix), respectively. Then in addition to (2.6) one has
also
<KxIISlE)IIK x ) = em/(ICx) = (1 + Ex - ilx.
E - ,Ix/2
)e i2Y /(KX). (2.7)
We also assume that all the other S-matrix elements are nonresonant, i.e.,
<KIISlE)IIK) = em/(IC) = e i2Y /(K) for K i= KR, Kx, (2.8)
where YlK) are the background phase shifts, which are functions of energy
E. One may be tempted to assume that the background phase shifts in the
resonant and non-resonant channels are slowly varying functions of the
energy. This is, however, in general not the case 2 and we shall not make
use of such an assumption here.
We shall call the above-described object a double resonance or double
multichannel resonance of quantum numbers (KR, Kx). Thus the formation
experiment in the scattering process
a+T->a'+T
can go two ways:
(a + T) withq.n'~A
~Rwithq.n'KR~('
~ ~ T + a withq.n.~'
')
--7
Xwithq.n.KX
2 These rapid energy variations follow from Wigner's eigenphase repulsion theorem: The
eigenphases <I/(K) do not cross (modulo n). As a consequence, near the resonance energy in
the channel KR the eigenphases in all other channels K must also vary. This is assured by the
existence of branch cuts in the individual eigenphases (and the corresponding transition matrix
elements <KII/}) which are also functions of E) which are in the complex energy plane much
closer to the real axis than the resonance pole. These branch cuts do not occur in the complete
S matrix and therefore do not have physical significance. The occurrence of these branch cuts
in the <KIi'1) and l'/(K) must be just such that all these branch cuts cancel out in the background
term (2.16), which can be a slowly varying function of energy. A discussion of these singularities
of the individual eigenphases is given in C. J. Goebel, K. W. McVoy, Phys. Rev. 164, 1932 (1967);
H. A. Weidenmuller, Phys. Letters 24B, 441 (1967); and R. H. Dalitz, R. G. Moorhouse, Proc.
Roy. Soc. A318, 279 (1970).
Figure 2.1
In the case of elastic scattering, '1 = '1A' (T' = T, a' = a); in the case of a
reaction, '1 = r ¥- '1k If one wants to illustrate this by a diagram analogous
to Figure XVIII.1.2, one would draw Figure 2.1.
The S-matrix for the double multichannel resonance is obtained by in-
serting (2.6), (2.1) and (2.8) into (2.3):3
<'1II SI(E)II'1') = ~ <'1IIK) <KII'1')ei2Y(K) + irR<~~K~ iK~II~:~;Y(KR)

+ ir x<'1IIKx) <Kxll'1')e i2Y (KX) (2.9)
Ex - E - ir xl2
The sum over K extends over all channels including the resonating channels
KR and Kx·
The amplitude for elastic scattering connected to the S matrix by (XVI. 1.48),
or (XVI.2.8), is obtained from (2.9) with '1 = '1' = '1A after a short calculation
using (2.1):
p71(p) = ei2Y(KR)(tr R 1<'1AIIKR)1 2 + tr xl <'1AIIKx) 12ei2 (Y(KX)-Y(KR»)

ER - E - irRI2 Ex - E - irxl2
+ L: <'1AIIK) <KII'1A)e iY (K) sin Y(K). (2.10)
K
In the case of inelastic scattering from the initial state with quantum numbers
rl' = '1A into the final state with quantum numbers '1 = r, the scattering
amplitude (XVI.1.48) is obtained from (2.9) as
r.:-=T(A~r)( ) = ei2Y(KR)(trR<rIIKR) <KRII'1A)

v' PrP r.l P ER - E - irRI2
+ tr x<rIIKx) <Kxll'1A)ei2 (Y(KX)-Y(KR)))

Ex - E - ir xl2
+ L <rIIK) <KII'1A)e i2Y (K)/(2i) (2.11)
K
Equations (2.9)-(2.11) are approximations to the same extent to which (2.6)

and (2.7) and (XVIII.6.23) are approximations. They have been derived here
from the ordinary Breit - Wigner approximation and the assumption that
the resonance (or quasistationary state) has internal quantum numbers K
incompatible with the internal quantum numbers '1 of the initial and final
state. Due to this derivation (2.9)-(2.11) contain the individual eigenphases
Y(K) though these eigenphases belonging to the intermediate quantum
3 We omit the index I in the phase shift 1, and 0, whenever possible. y then always denotes the
background phase of that partial wave in which the resonance occurs.
numbers K may be unphysical quantities. The physically observable quantities

are the resonance terms and the total background term (summed over all K)
and only these quantities should be compared with the experimental data.
XX.2a Single Multichannel Resonance

Before we continue the discussion of a double resonance, we treat the simpler
case that there is only one resonance with quantum number K R . Such a
resonance we shall call a single multichannel resonance. This case is obtained
from the above formulas (2.10) and (2.11) by omitting the second term in
the bracket.
It is customary to introduce the following notation:
(2.12)
At this point we use time-reversal in variance. If time-reversal invariance
holds, then the unitary matrix <I]IIK) is orthogonal, i.e., the matrix elements
are real [Equation (2.1a)]. Thus for time-reversal-invariant interactions
r~/2 are rea\.4
Because of (2.1), the r ~ fulfill
L r~ = r = r R • (2.13)
r ~ is called the partial width for the channell]. With this notation the elastic
scattering amplitude for a single resonance of quantum number KR is
written
PT.(p) = ei2Y(KR) r~A/2 + b(n ) (2.14)

I ER - E- ir/2 ., A ,
and the reaction amplitude is given by:
(2.15)
where
(2.16A)
K
and
(2.16r)
K
which can be a slowly varying function of energy even if the individual Y(K)
vary rapidly. The representation (2.14), (2.15) is called the generalized
Breit ~ Wigner approximation. These expressions are very similar to the
corresponding expression (XVIII.6.26) for the one-channel resonance, ex-
cept that the height of the multichannel resonance is smaller by the factor
4 If time-reversal invariance does not hold, r~!2 is complex and one defines r. by r. = Ir~!212
We shall not discuss this case here.
<I]AIIKR)2 or <I]AIIKR) <rIIKR)' The first factor represents the transition am-
plitude (square root of the transition probability) for a transition from a
state with quantum numbers I]A to a state with quantum numbers KR and
back to a state with quantum numbers I]A; the second factor represents the
transition amplitude from I]A to KR and from KR to r. Thus the smaller height
is due to the inelasticity, i.e., the fact that the resonance contributes not only
to the channell]A or r, but also to all the other channels I] =f. I]A or I] =f. r.
The cross sections corresponding to the amplitudes (2.15) will then not
look much different from the expressions for the cross sections given in
Sections XVIII.6 and XVIII.9. Using the abbreviations:
ER - E
(2.17)
f = rj2 '
(2.18)
where KR are the quantum numbers of the resonance R, and I]A the quantum
numbers of the initial state A = (a, T), one can write the lth partial cross
section (Jt . . ~) using (2.14) for I] = I]A and (2.15) for I] = r with (2.16):
U~A""~A) = 4n(21 + 1 \ ~ <I]AIIK) <KIII]A)e'Y(K)

1) p2 . sin Y(K) _ i \2
+ e,2. 1'(KR) fR~A
(2.l9A)
For I] = r one obtains from (2.15) with (2.l6r):
1 \I ei2Y (K)
+ e 2iY (KR) _R_•\2•
r
U\A .... r) = 4n(21 + 1)2 <r II K)<KIII]A)-2'- (2.19r)
p K i f - 1
After some calculation this can be written as
UjA .... ~A) = 4n(21 + 1) 12 [lb(I]A) 12 + 2R~Al (R~A + I <'lAIIK)<KII'lA)

+
J
p f K
x sin Y(K)(2f coS{2(Y(KR) - Y(K»} - 2 sin{2(Y(KR) - Y(K»}»)
(2.20A)
+ 1)2 R - ( R' + I
1 [ Ib(r)1 2 + - 2
r
u~A .... r) = 4n(21 <rIIK)<KII'lA)
+1
1
p f K
X (-f sin{2(Y(KR) - Y(K»} - coS{2(Y(KR) - Y(K»}») (2.20r)
where b('lA) and b(r) are given by (2.16). Equation (2.20A) is the elastic
cross section and (2.20r) is the reaction cross section.
For an energy E far away from the resonance energy E R , i.e., for lEI ~ 0,
the second term in the bracket on the right-hand side of (2.20) is very small
and vanishes for If I ~ 00. The cross section is then given by
U~A""~) ~ u(nr) = 4n(21 + 1) ~ Ib('lW for If I ~ 00. (2.21)

p
Thus we see that (2.20) is very similar to (XVIII.9.14): a single resonance in

the many-channel case will appear as a structure above a gradually varying
background. The shape of this structure will in general not be the same as
the one given in Figure XVIII.9.2 and will depend upon the various back-
ground phase shifts y. However, if these background phase shifts are, in the
neighborhood of the resonance energy E :::::: E R, approximately the same
modulo n for all quantum numbers K:
yz{K) = yz{E) = y(E), (2.22)
then one can write (using <'1AI'1A) = LK<'1AIIh:)<KII'1A) = I),
(A~q)
(J/
--->
(J(nr)
+ Rq2 + [22q[+- 1 2 + R R(J(bg)
/
l"
lor '1 = '1A'
(223)
.
where q is again defined by

q = cot yz{E) (2.24)
and (Jjb g ) is given by
(J~bg) = 4n(21 + 1) ~ sin 2 y. (2.25)

P
Equation (2.23) is, indeed, very similar to (XVIlI.9.14) with (XVIII.9.6).
The shape of a single resonance in the many-channel case will, therefore,
be very similar though not identical to the shape depicted in Figure
XVIII.9.2.
From (2.20) we see that the multichannel resonance shows up in the
elastic channel, I] = I] A, as well as in all the reaction channels '1 = r, at the
same energy ER and with the same width r. The shape of the resonance
depends upon the background phase shifts y(K). And even if (2.22) holds,
the resonance with quantum number KR will have different though similar
shapes in the various channels 1], because R in (2.23) depends in general
upon '1 according to (2.18).
However, if one has the particular case that the background for all quan-
tum numbers K-the reasonating K = KR and all the nonresonating K-is
negligible, then one has
(2.26)
Thus, the resonance without background appears in all channels as a Breit-

Wigner resonance at the same energy and with the same width, but with
different heights depending upon the value of the partial width rn'
Further, as '1A can really take any value, the same Breit-Wigner resonance
also appears for any initial state. Thus, independent of what the initial and
final states of a scattering process are, if only their transition amplitudes to
the state with quantum numbers KR are different from zero, the resonance will
appear at the same energy and with the same width. This is, of course, what
one must expect of a quasistationary state, which is entirely determined by its

"quantum numbers" (ER' r, 1, K R ) and independent of the experimental
conditions. 5
Experimental examples of single multichannel resonances are shown in
Figure XIX.3.l. This figure shows the various resonances (" energy levels")
of Sj28 as they occur in the processes
A: + P --+ (Si 28 )* --+ P + AF7,

AF7
B: Al27 + p --+ (Si 28 )* -+ + Mg24, !y.
c: Mg24 + -+ (Si 28 )* -+ p + A1 27 ,
!y.
D: Mg24 + -+ (Si 28 )* -+ + Mg24.

!y. !y.
(Si 28 )-resonances are observed at the following energy values: 6 2.75, 2.93,
2.95,3.00,3.14,3.20,3.28 MeV. Fine discrepancies between the curves-like
the double peak around 3.4 MeV in process B, which is matched only by a
single peak in procees C-can be explained in terms of differences in the
energy resolution for the experiments.
Let 1JA denote the quantum numbers of the p_AJ27 system, r the quantum
numbers of !Y._Mg24, and KR1' KR2' ..• the quantum numbers of the various
(Sj28)-resonances. The energy values of these resonances are sufficiently far
apart that one can consider each separately as a single resonance and need
not apply the considerations for a double resonance described below.
We see that in general, each single resonance occurs in all four processes.
However, whereas in the reactions Band C they usually occur as pure Breit-
Wigner resonance, in the elastic process A and in the process D they have con-
siderable background phases. For example, the resonance at 3.203 MeV,
whose quantum number we may call K R6 , occurs in Band C as a bump and in
A and D as a dip. This indicates that <'1 A I KR.> as well as <r I KR.> are different
from zero, and that the sum over K in (2.20) is negligible for '1 = r but not
negligible for '1 = '1A = (quantum numbers of the p_A1 27 system) and
'1 = r = (quantum numbers of the !Y._Mg24 system). A possible explana-
tion for this effect is that the background phases Y(K) are not negligible,
but approximately independent of K, so that for the elastic processes some-
thing like (2.23) holds. Under the same assumption one obtains for the
inelastic processes, using <'11 '1A> = 0,
(2.23')
As we have remarked already in Section XVIII.6, the notion of the quasistationary state
5
r, /, K R ) has a meaning only up to a certain accuracy. If the energy dependence of r has
(E R ,
to be taken into account, then the apparent position, E\lbmved), may be different for different
experiments.
" These are the energy values on the IX-particle energy scale. The difference in the origin of the
energy axis for the IX-particle energy and the proton energy is 1.61 MeV, and the energy scales
of the processes A, Band C, D have been adjusted by bringing each pair of resonance peaks of
the reaction curves Band C into coincidence.
so that in a reaction the resonance occurs as a pure Breit - Wigner above a

possible background (j(nr).
XX.2b Double Multichannel Resonance
We now return to the case of two resonating lth partial waves: one with the
internal quantum numbers KR and energy and width (ER' r R), and the other
with internal quantum numbers Kx and resonance parameters (Ex, r x). We
start from (2.10) and (2.11) and consider simultaneously the cases of elastic
scattering of (2.10) IJ = IJA and of the reaction (2.11), 1'/ = r. We use the
notation
R = R~ = <IJIIKR) <KRIIIJA),
X = X~ = <IJIIKx) <KxIIIJA)ei2 (Y(KX)-Y(KR)). (2.27)
We shall always assume time-reversal in variance (2.1b), so that R is real
and the phase of X is entirely determined by the difference of the back-
ground phases. Equations (2.10) and (2.11) are then written
T,( ) = ei2Y(KR) ( (r R/2)R~ + (r x/2)X~ ) +b (2.28)

PIP ER - E - irR/2 Ex - E - irx/2 '
where b is given by (2.16).
To simplify our following discussion we ignore the background term b.
It will have effects similar to those discussed above for the single resonance
and in Section XVIII.9.
The cross section for the double multichannel resonance then becomes
(if we assume for the sake of simplicity that X is real):
(j(R)
I
= 4n(21
+
1) ~ (n/
p2 (ER - E)2
4)R 2
+ n/4 + (Ex
Cri/4)X2
- E)2 + ri/4
(rR/4)r x2RX[(ER - E)(E x - E) +crR/4)r x])
+ [(E R - E)2 + (n/4)][(E x - E)2 + (ri/4)]· (2.29)
Thus the lth partial cross section is the sum of two Breit-Wigner cross
sections, with different widths and different heights, and a third term (the
interference term). A double resonance is characterized by four parameters
(ER' r R ; Ex, rx) and the quantities R and X, which depend according to
their definition (2.27) upon the final-state quantum numbers IJ and also
upon the initial-state quantum numbers I'/A. Thus for one channel R and X
may have the same sign, and for another channel they may have opposite
signs. The shape of the cross section of a double resonance, therefore, varies
with the channel quantum number. If in one channel it appears as a double
hump, it may show up in another channel as one single broad hump. The
most favorable circumstances for a double hump are R ~ X. Then the third
term is (2.29) has two maxima, one near E = ER and the other near E = Ex,
which enhance the humps at these values that come from the two Breit-
Wigner cross sections in (2.29). Then the double resonance will appear
in the cross section as two maxima with different widths, provided the back-
ground can be ignored.
To obtain an idea of the possible shapes of such a double resonance, let
us consider some special cases. It is useful to introduce the new variable
(2.30)
and denote
~ = ER - Ex· (2.31)
Then the cross section formula (2.29) is written as
aiR) _ 4n(21 1) ~ ! (R2n x 2 rl

I - + p2 4 (~/2 _ g)2 + (rR/2)2 + (-N2 _ g)2 + (r x/2)2
2RX(g2 - (~/2)2 + r Rr x/4)rR r X )
+ [(~/2 - g)2 + (r R/2)2] [( - N2 - g)2 + (r x/2?] . (2.32)
The interesting situation arises when ER is close but not identical to Ex and
if the widths of the two resonances have approximately the same magnitude
as ~ = ER - Ex. Therefore we shall consider the case
(2.33 1 )
Note that in this case the third term in (2.32) has a zero and a minimum
(if sign R = sign X) or maximum (if sign R = - sign X). With (2.33) one
obtains from (2.32)
(R) _ 1 r2 ( R2
al - 4n(21 + 1) p2 4 (N2 _ g)2 + (~/2)2
x2 2RXg 2 ) (2.34)
+ (-~/2 - g)2 + (N2)2 + g4 + (~2/2)2 .
The term in the bracket of (2.34)-i.e., aIR) except for the factor
4n(21 + 1)2"-
1 r2
p 4
-is depicted in Figure 2.2 for various values of X (R = 1). Ignoring
ei2 (Y(KR) - Y(KX)), X is connected to the coupling constants according to (2.27) by
X = <I1IIKx><KxIII1A>.
<11 II KR> <KR III1A>
The correspondence between the curves and the values of X is as given in
the legend. We see that the cross section may look quite different for dif-
ferent couplings between the initial and final channels and the resonance
channel. For X = lone obtains a double peak with the maxima slightly
&= - 3~ & = -~ £ = 0 & =~ & = 3~
Figure 2.2 Cross section for double resonance: the case (2.33 1 ) for various values of X:
Symbol: x + \l f:::. o
X: 1.0 0.5 0.0 -0.5 -1.0
below E = Ex (g = -d) and slightly above E = ER (g = +d). For

X = -1, the cross section has one single broad bump centered around
E = !(E R + Ex). For X = 0 there is, of course, only one bump at E = E R,
representing the resonance R, as in this case the resonance X does not couple
to the channels IJ or IJA' But also for X = - 0.5 there is only one single
bump, which however does not have the form of a Breit-Wigner resonance.
Thus, for different final channels IJ, a double resonance may show up in the
cross section as a double peak, as a broad single peak with maximum between
the two resonance energies, or as a single peak around one of the resonance
energies.
The cross sections in Figure 2.2 correspond to the most symmetrical case,
(2.33 1 ), If r x or r R differ from d the picture changes slightly. Figure 2.3
depicts (2.32) [except for the factor 4n(21 + 1)(1 / p2)r2] for the case
rx = d,
The variation in the shape of the cross section with the coupling between
the channels is similar to that in Figure 2.2.
In the above considerations we have ignored the background term b, so
that the above discussions are valid only if (2.16) vanishes. Also we have
ignored Y(Kx) - Y(KR)' From our discussions in Section XVIII.9, and from
the footnote on p. 520 we know already that the background phase shifts
may be large, so that this idealization may well be inadmissable. Taking the
interference between resonances into account will make the complicated
features of a double resonance even more complicated. Taking in addition
c= - 3~ t = -~ t = 0 c +~ c= 3~
Figure 2.3 Cross section for double resonance: the case (2.33 2 ), Symbols as in Figure
2.2.
the background from other partial waves and the resolution of the experi-
mental apparatus into account, one concludes that the observation of a
double multichannel resonance as a double peak in the cross section is an
unlikely possibility. Thus two resonances with different quantum numbers KR
and Kx but with energies ER and Ex, that differ from each other by an amount
comparable with the width and that both couple to the initial and final
channel, will be very hard to observe.
Nonetheless, examples of double resonances have been seen in nuclear
scattering processes. The most prominent examples are the Be 8 energy
levels with (angular momentum)parit y = t = 2 +. These resonances have
been observed in the processes
He 4 + He 4 ---> (Be 8 )* ---> He 4 + He 4 (2.35 1 )
and
He 4 + He4 ---> (Be 8 )* ---> Be 8 + y, (2.35 2 )
at excitation energies of 16.6 and 16.9 MeV of the Be 8 , with widths of ap-
proximately 110 and 80 keV, respectively. We do not want to describe the
details of this system, and shall just establish the connection with the quanti-
ties of our formulas (2.28) and (2.29). We ignore the background phases;
then (2.28) can be written
PT I
A~~A _ !rx<l1AIIKx> <KxllllA>
-
+ !rR<l1AIIKR> <KRllllA> (2.3 6)1
Ex - E - ;r x/2 ER - E - ;r R/2
and
TA~~ = !rx<l1IIKx> <KxllllA> + !rR<l1IIKE> <KRllllA>
P I Ex - E - ;r x/2 ER - E - ;r R/2 .
'" 800
C
::I
8 600
.,
-0
400
U
~
0 200
u
800
600
400
200
17.42 17.22 17.02 16.82 16.62 16.42 16.22

Excitation energy in MeV of Be"
Figure 2.4 Fits of the doublet structure in Be 8 in the He 4 _He 4 channel, Note that the
energy decreases from left to right. [From W. D. Callender et al. ~ with permission.]
Here I17A) denotes the internal state of the (He 4 , He 4 ) system, 117) denotes
the state of the (Be B , )I) system, and 1 Kx) and 1 K R ) denote the internal states
of the Be B * (16 MeV) and Be B * (16.9 MeV), respectively, so that (2.36J is
the scattering amplitude for the process (2.35;). Depending upon the ratio
(17A I/ Kx )z/ (17AI IKR / , the cross section for the process (2.35 1 ) will then have
the shape of one of the curves in Figure 2.2 or 2.3. The cross sections have
been measured 7,B and the experimental data points together with a fit to
the data are plotted in Figure 2.4. B
The curve in the upper plot shows a fit to the data obtained with an ex-
pression (2.29) for a double resonance. The lower plot illustrates the fit
with the sum of two Breit - Wigner cross sections,
Al Az
(2.37)
(J ~ (Ex _ E)2 + (rx/2)2 + (ER - E)2 + (rR/2?'
7 A. D. Bacher, F. G. Resmini, H. E. Conzett, R. deSwinairski, H. Meiner, 1. Ernst, Phys. Rev.

Letters 29, 1331 (1972).
8 W. D. Callender, C. P. Browne, Phys. Rev. 2, 1 (1970). This figure does not give the cross
section for the formation process (2.35,) but for the production process
B'o + d -+ IX + Be 8 *
L IX + IX.
R and X in (2.29) are therefore not
but
with
6
-..
!!l
·2 5
~
:e'"
!:: 4
~
"0
3
0:;
> 2
O~-=~~-L--~--~--~--~--~--~
16.5 16.6 16.7 16.8 16.9 17.0
Excitation energy in e or Be 8
Figure 2.5 Fits of the double structure in Be 8 in the Be 8 -1' channel. [From Nathan
et al.9 • Phys. Rev. Lett. 35, 1137 (1975), with permission.]
which would hold if the states at 16.6 and 16.9 MeV were unrelated single
resonances. This fit with two noninterfering Breit - Wigner curves is clearly
unacceptable, whereas the fit in the upper plot with a double resonance (two
interfering Breit-Wigner curves) is excellent. Thus we conclude that the
double-peak structure observed around 16.6 and 16.9 MeV of the Be 8 *
system is a double resonance. Both bumps have approximately the same
height, the bump at 16.9 MeV is slightly narrower, and the fitted curve re-
sembles the curve (marked x) in Figure 2.3 for XjR ~ 1. Both states IKR>
and IKx> couple, therefore, with approximately the same strength to the
incident channel:
(2.38)
Figure 2.5 shows a plot of the experimental data 10 for the reaction (2.35 2 ) ,
The solid line is a fit with (2.29); the dashed line is a fit with (2.37). Again the
sum of two noninterfering Breit - Wigner curves is clearly ruled out, and a
double resonance gives a very good fit. This time, however the curve re-
sembles more the curve (marked +) in Figure 2.3 for XjR ~ 0.5. From
this and (2.38) we conclude that the coupling of the two IKR> and IKx> to
the outgoing channel (Be 8 , }') fulfills the relation
(2.39)
The experimental data for the Be 8 system around 16.6 and 16.9 MeV give
a beautiful illustration of the fact that a double resonance may show up in
the cross section with a variety of shapes.
9 A. M. Nathan, G. T. Garvey, P. Paul, E. K. Warburton, Phys. Rev. Lett. 35, 1137 (1975).
XX.3 Argand Diagrams for Inelastic Resonances
In Section XVIII.8 we discussed the behavior of p7;(p) in the Argand diagram

for the case of purely elastic scattering. In the second part of that section
we also described how 7;(p) can be obtained from the experimental data.
In the present section we shall give a brief discussion of the behavior of the
elastic partial-wave amplitude p7;(p) in the presence of inelastic processes
and of the partial-wave reaction amplitude JP:P
T,jp). The determination
of these amplitudes from the experimental data follows essentially the same
procedure as described in the second part of the Section XVIII.8: for the
determination of 7;(p) one uses the elastic differential cross sections, and for
T,jp) the differential cross section for the reaction (initial state I} A) ~
(final state r).
We will now assume that there is a single multichannel resonance and
discuss how this single resonance will show up in the Argand diagrams for
the elastic scattering amplitude for the reaction amplitude. We first treat
the idealized case without background. The scattering amplitudes are then
given according to (2.14) and (2.15) by
T(R)( ) _ ~r~A (3.1)

p 1 P - ER - E - if'/2'
1 r1/2r1/2
r.:-::. T(R)( ) _ 2" r ~A (3.2)
V PrP r,l P - ER - E - if'/2'
Here P is the initial momentum and Pr is the final momentum, which is a

function of the initial momentum and the difference of internal energies in
the initial and final states. 1 0
rr and r ~A are the partial widths (2.12) for the reaction channel r and for
the initial channell} A' respectively. ERand r R = r are the energy and total
width of the resonance. We use again the abbreviation
2
E = (ER - E)- (3.3)
r
and also define
(3.4)
~ may be negative.
Equations (3.1) and (3.2) can then be written
PT(R)
1
= x nA
., (3.5)
E-l
(3.6)
1" For a massive nonrelativistic projectile of mass m A before and mass mb after the reaction,
p, = Pb is given by the formula (XVI. 1.24).
Im(pT) Im(pT)
-t o Re(pT) -t o Re(pT) !
(a) (b)
Figure 3.1 The resonant amplitude for the elastic channel is shown for two different
values of the elasticity X~A: (a) X~A = 0.75 and the phase shift for the amplitude goes
through 90° at the resonant energy; (b) X~A = 0.4 and (j = 0° at the resonant energy.
Notice that the phase of the resonant amplitude in the resonance circle goes through
90° in both cases. [From Barbaro-Galtieri (1968), with permission.]
X~A = <'1A.IIKR)2 is called the elasticity of the resonance and is the transi-
tion probability for a transition from a state with quantum numbers KR into
a state with quantum numbers '1A at a particular value of E. Thus X~A rep-
resents the "fraction" of the resonance with internal quantum numbers KR
that is coupled to the elastic channel '1A' For X~A = 1 the resonance couples
to the elastic channel only, all x~ with '1 #- '1A are zero, and pTI R) describes
the unitary circle for an elastic resonance as described in Section XVIII.8.
The elasticity X~A is related to the inelasticity coefficient '11(E) defined in
(XVI.2.l0). At the resonance energy this relation is '11(E R) = 2X~A - 1. In
contrast to '1lE), which has a strong energy dependence, empirical data
show that the elasticity X~A is only very weakly energy-dependent and is
therefore a more convenient parameter.
The Argand diagram of (3.5) is depicted in Figure 3.1. 11 It is a circle of
diameter X~A lying inside the unitary circle. The resonance energy, f = 0,
lies on the top of the circle, and the points f = =+= 1 correspond to E =
ER ± r /2. One has to distinguish two cases:
(a) X~A:;::: 0.5; then the resonance occurs at a phase-shift value

JI~A)(ER) = n12.
(b) X~A < 0.5; then the phase shift at resonance is JI~A)(ER) = O.
However, there is no qualitative difference between these two cases; in both

cases the resonant eigen-phase-shift J(KR)(E) goes through nl2 at E = E R.
11 Figures 3.1, 3.2, and 3.3 are from Barbaro-Galtieri (1968).

Im(pT)
+;
£=0
£ = -1 e = +1
-; +t Re(pT)
-t
Figure 3.2 The resonant amplitude for scattering into an inelastic channel. [From
Barbaro-Galtieri (1968), with permission.]
The Argand diagram for the resonant reaction amplitude (3.6) is depicted
in Figure 3.2. For the diameter of the circle one takes 1 xrx~A I. It lies J
inside the circle of Figure XVI.3.2 for the reaction amplitude, and
If the background is included, the elastic scattering amplitude (2.14) is

written
pI; = e i2 y X~A. + b. (3.7)
f-l
Im(pT)
-; t Re(pT)
Figure 3.3 Elastic scattering amplitude for the case of a resonance and a background
amplitude. If the background is assumed to be constant throughout the resonance
region, the amplitude pT is on the circle with center C. In this figure the background
amplitude is not purely elastic, but represents some absorption . [From Barbaro-
Galtieri (1968), with permission.]
The graphical representation of (3.7) for y = 0 is obtained from the

graphical representation of (3.5) in Figure 3.1 by displacing the circle by
the vector b and starting the resonance circle at the endpoint B of b. The
resonance energy is then at the top of this resonance circle. The effect of the
background phase shift y is then a rotation of the resonance circle by the
angle 2y. The resulting Argand diagram for (3.7) is depicted in Figure 3.3.
The Argand diagrams depicted in the figures of this section and also in the
figures of Section XVIII.8 are idealizations. The energy dependence of the
background beE), the background phase shift y(E), and the elasticity dis-
torts the circular behavior of the amplitude. An example of an experimental
Argand diagram, obtained from the experimental data by a procedure
described in the second part of Section XVIII.8, is depicted in Figure 3.4
below.
The existence and nature of a resonance in the Argand diagram are then
determined by comparison with the idealized cases depicted in the above
figures. If there is an indication of a circle, one takes this as a sign for a
resonance. The resonance energy ER is then determined as that value of E
Im(pT) Im(pT)
Re(pT)
( -1680) 6 (1237)
Im(pT)
(-15SOand 1715)
Figure 3.4 Experimental Argand diagrams for some pion-nucleon partial-wave
amplitudes. [From A. H. Rosenfeld et al., Data/or Elementary Particle Physics, 1965,
with permission.]
at which p7;(E) and the phase shift c5 1(E) change most rapidly. Around ER
the path pTz(E) may show as a segment of a circle, and the elasticity X~A is
then obtained as the radius of this circle. The total width r may then be
obtained from the energy difference between the points i circle to the left
and i circle to the right of the resonance energy on the circle segment. In
this way Argand diagrams have been of great help for displaying resonance
behavior and for determining the resonance parameters.
As an illustration, we depict in Figure 3.4 Argand diagrams obtained from
the experimental data for some partial-wave amplitudes in 1T-P elastic
scatteringY Each diagram shows the unitary circle and, inside it, pTz(E)
for various partial waves. (For 1T-P scattering there are several partial waves
for each value of I, characterized by isospin I and total angular momentum
j = 1 ± i; we cannot discuss these things here, they are also of no relevance
for our present discussion.) The numbers at the curves for pT(E) indicate
the energy of the 1T-P system in Me V = 106 eV. The second diagram (labeled
by P33) shows the almost purely elastic resonance ~ at the energy 1236 MeV.
The Argand diagram is almost exactly as described in Figure XVIII.8.1.
The third diagram (labeled S 11) is the Argand diagram of another partial
wave with two resonances, one around 1550 MeV and the other around
1715 MeV, both inelastic. The first diagram (labeled S31) shows another
case of an inelastic resonance around 1680 MeV. We see from these examples
that the behavior of different partial-wave amplitudes is quite distinct and
in general far from the idealized behavior. The partial wave in the second
diagram, P 33, is in fact an exception; all other resonances are inelastic and
have a large background term.
12 7! stands for 7!-meson and p for proton.

CHAPTER XXI
The Decay of Unstable Physical

Systems
In Section XXI.1 the decaying state is introduced as a resonance for which

the production process is ignored. Section XXI.2 gives a heuristic discus-
sion of decay probability (decay rate) and its measurement. Section XXI.3
describes a decaying state by a generalized eigenvector of a Hermitian energy
operator with complex eigenvalue. Using this novel concept, the calculation
of the decay rate is very simple and is given in Section XXI.5. Section XXI.6
contains a discussion of the partial decay rates and the use of various basis
systems for their calculation.
XXI.l Introduction
In Chapters XVIII and XX we discussed scattering processes in which an
unstable physical system was formed as a quasistationary intermediate
state. The measurements were performed at the initial state, which was
prepared in the distant past, and at the final state, which was observed when
the interaction was no longer in effect and the quasistationary state had
ceased to exist. We have already seen in Chapter XX that the property of
the quasistationary state is-at least to a high degree of accuracy-indepen-
dent of the process in which it is observed. Its characteristics are the energy
E R , the width r, and the internal quantum numbers /CR.
The formation process
a+ T--+R--+a' + T' (1.1)
537
538 XXI The Decay of Unstable Physical Systems
a a'
Initial state Final state
with quantum Resonance with E R , r, KR with quantum
number·'1A number r
T'
Figure 1.1 Resonance scattering.
in which the quasistationary state is detected is depicted by the diagram in

Figure 1.1, and (ignoring background) is described according to (XX.2.1S)
by the scattering amplitude
r::-: r/2 (1.2)
vi PrpT,.,1 = <rIKR> ER _ E _ ir/2 <KRI'1A>'
If r is sufficiently small and the delay time according to (XVIII.6.36) and
(XVIII.6.43) sufficiently large, one may ignore the formation process. One
then starts the description at a given time, t = 0, with a physical state
Wet = 0) = WR, and describes only that part of the process of Figure 1.1
which is depicted by Figure 1.2(a). R is then considered as a decaying physical
system or a decaying state of a physical system. Figure 1.2(b) depicts the
earlier part of the resonance scattering process terminating at t = O. Here R
is considered as a forming state.
Examples of decaying states are numerous in physics; "real" stable states
are very rare. The radiative transitions of excited atoms or molecules A *
to lower states A,
A* ~A + 'Y, (l.3a)
are typical examples of decay processes. The unstable physical system had,
of course, first to be formed, and this may have happened by a process like
'Y +A ~ A*. (l.3b)
But for the excited states of atoms one ignores the mode of formation and
starts with an initial state W R describing the decaying system A *.
The crucial problem is thus the choice of W R at the arbitrarily chosen
initial time t = O. From WR, if the time development is described by a
Hamiltonian H according to the basic assumption V of Chapter XII, one
can then obtain the state of the decaying system at any later time t by
(1.4)
a
r
R withER - i -
2
T
(a) (b)
Figure 1.2 (a) Decaying state; (b) forming state.
XXI.2 Lifetime and Decay Rate 539
Equation (1.4) is valid if the decaying system is isolated and evolves un-
disturbed. Whether for a realistic decay process the decaying system really
evolves undisturbed is questionable, because it almost unavoidably inter-
acts with its surroundings, especially when the decay process is observed. 1
Therefore the decaying system may be subjected to measurement processes
occurring at random times. For example, a decaying particle in a bubble
chamber leaves a track of bubbles caused by the interaction of the decaying
particle with the environment. Each bubble means a measurement in which
the system has been found undecayed. Such a measurement at time tl
changes the state, according to the basic assumption IIIb, into the state
(1.5)
where Ab is the projection operator on the space of states of the undecayed
system or a subspace of it. Thus in addition to the change of state (1.4) one
has to consider the randomly occurring change of state (1.5). and the ob-
served lifetime will not be the same as the lifetime of the unobserved decaying
state developing according to (1.4).
All these problems we shall not discuss here. 2 We shall treat the idealized
situation where the decaying state can be considered as an isolated physical
system whose lifetime is not affected by a measurement process (1.5) caused
by the environment.
XXI.2 Lifetime and Decay Rate
In a decay experiment one has an ensemble of unstable physical systems R

in a volume "Y, which is surrounded by counters that detect the decay
products a'. The counting rate, i.e., the number of decays per second, NIT,
is proportional to the number of unstable particles, N R = PR "f/. The pro-
portionality constant
A = NIT (2.1)
NR
is called the initial decay rate, or often just decay rate.
The number of decays per second, NIT, is equal to the rate of decrease
of the number of unstable particles R, dN Rldt. Therefore, by (2.1),
dN
- dtR (t) = AN R(t), (2.2)
1 A Beskow. J. S. Nilsson. Arkiv/or Fysik 34,561 (1967).

2 For a review of these subjects the reader is referred to L. Fonda. G. C. Ghirardi. and
A. Rimini. Decay theory of unstable quantum systems. Reports on Progress in Physics 41, 587
(1978). Further methods not discussed in this review can be found in A. P. Grecos: Decaying
states in quantum systems. Singularities and Dynamical Systems. edited by F. N. Pnevmatikos.
North Holland, Amsterdam (1985) and A. George, F. Henin, F. Mayne, I. Prigogine, New
quantum rules for dissipative systems. Hadronic Journal 1, 520 (1978).
at least at t = O. And if there is no further source or sink of unstable particles

R in 1/, one is led from (2.2) to the exponential decay law
(2.3)
where N R(O) is the number of unstable particles R at the time t = 0 when

the decay process started.
As N R(O) = - J~o dN R(t), it follows that -dN R(t)/N R(O) is the proba-
bility of a decay at time t during the time interval dt. The decay rate, i.e.,
the probability per unit time for a decay at t, is therefore
d[l} == &(t) = _ ~1_ dN R(t2. (2.4)

dt NR(O) dt
&(t)lt=o = A, (2.5)
which explains the name "initial decay rate" for A.
The average lifetime of an unstable particle R,
r = 100
tg7>(t) dt, (2.6)
is called the mean life, or just the lifetime. If (2.3) holds, then
r = f
oo
0
1 - At
N R(O) N R(O)e At dt
1
= 1 = &(t =
1
0)' (2.7)
In quantum physics the decaying particles are quantum physical systems

and the observed quantities are probabilities. The observables measured on
the decaying state Wet) may be the "property" of being in the original
resonance state W R. The expectation value of W R in the state Wet) =
e- iHtWReiHt represents the probability that R has not decayed. It is called
the nondecay probability:
(2.3a)
It corresponds to the observed quantity N R(t)/N R(O).
Another observable that may be measured is the projection operator I\.
on the space of physical states of the decay products (at, T'), i.e., the property
I\.. The expectation value of I\. in the state Wet) represents the probability
for observing at and T'. This transition probability from the state Wet)
to 1\.,
[l}(t) = <I\.)t = Tr(I\.W(t)), (2.8)
is called the decay probability of R into (at T'). Obviously, if R --> at + T'
is the only decay possible, then the sum of the probability that R has not
decayed and the probability that R has decayed into a' + T' must be one:
.?JlR(t) + [l}(t) = 1. (2.9)

XXl.2 Lifetime and Decay Rate 541
If R has several decay channels, i.e., if it can decay into several decay
products labeled by 11, whose spaces of physical states are A~ X, 11 = 1,2,3, ... ,
with A~A~. = 0 for 11 -1= 11', then
A == At + A2 + A3 + ...
and the probability for a decay into any of the channels 11 is
.9J,,(t) = Tr(A~ Wet»~. (2.10)
(2.9) is also valid for this transition probability and states that at any
arbitrary time t either R has decayed or has not decayed.
The quantum-mechanical decay rate is the transition rate from Wet)
into A:
d& . d
dt = &(t) = dt (A)t· (2.11)
In (XIV.2.18) we have already calculated an expression for it, which we shall

use below.
The experimentally measured quantity is the lifetime t or the decay rate
A.. For large lifetimes (from 1 sec to 1 year), one measures N R(t) as a function
of t and obtains A. from (2.3). For shorter lifetimes down to 10- 9 sec one
measures the lifetime by electronic methods. The decay to be studied must
be preceded by another event, e.g., the formation of the unstable system,
which is used to establish the origin of time. A counter is activated by this
event and remains activated for a time t. The number of times the activated
counter detects a decay divided by the number of times the counter has been
activated is the observed quantity which is proportional to the probability
that a decay has occurred during the time interval between 0 and t. This is
according to (2.4) and (2.3) given by
Thus
Number of times the activated
k r dt' A.e- At• = k(l _ e-.l.t) = counter detects a decay
Jo Number of times the counter is
activated
(2.12)
One measures the ratio on the right-hand side as a function of t. As k is a
constant independent of t [connected with N R(O) and the detector efficiencyJ,
one can calculate A. if one knows (2.12) for several values of t.
For lifetimes shorter than 10- 9 sec one cannot directly measure t, but
has to resort to indirect methods. As we shall see below, the decay rate is
equal to the width. One then obtains the lifetime from the measurement of
the width in resonance scattering.
XXI.3 The Description of a Decaying State and the

Exponential Decay Law
In Section XVIII.6 we have discussed resonances, quasistationary states,

Breit - Wigner amplitudes, and resonance poles of the S-matrix. The con-
nection is summarized in the diagram of Figure 3.1. A pair of second sheet
poles of the S-matrix is the property that implies the three other properties
listed above it in the diagram. Whereas a resonance is always a quasi-
stationary state and vice versa, these two properties lead to a Breit - Wigner
amplitude only if the fourth and higher derivatives of (jz{E) are small. And a
Breit-Wigner amplitude does not necessarily imply the existence of an S-
matrix with a pair of second sheet poles, as the S-matrix entails further
properties like analyticity (causality) and unitarity. We will, therefore, take
the pair of second sheet poles of the S-matrix near the real energy axis as the
definition of a quasistationary state. From this definition we will deduce in
the next section the vectors that describe decaying states, which we will call
Gamow vectors or Gamow kets. In the present section we will introduce the
decaying states more heuristically and state some of their properties, which
will be derived in the next section. These properties will be used to calculate
the decay probability and in Section XXI.5, the decay rate.
The decaying state W R can be prepared by a scattering experiment in
which the time delay is large. One obtains the decaying state if one isolates
the intermediate quasistationary system, ignores its mode of formation and
considers it as an initial state which at time t = 0 starts the time development
which leads to the decay. As the scattering amplitude of a resonance scatter-
ing process has the Breit-Wigner form, we expect the decaying state W(O) =
W R to have a Breit-Wigner energy distribution.
For the sake of definiteness we will assume that the spectra of Hand K
are continuous and go from 0 to 00. This is not a serious restriction of
R'r:(~:~'~d:;~::'PidIY i"""i'g J,(E))
QTista~i,;~~r~ :~:~~E~~filn: by a l~~e value of the time delay t~)(E))

I ER-E-ir/2
dn(j(E)
only if ~ is small for 11 ~ 4
1
Breit-Wigner Amplitude (Siegert pole)
r
only if the S-matrix is analytic (causality) and unitary
Pair of Second Sheet Poles of the S-matrix
Figure 3.1 Relations between a resonance, quasistationary state, Breit-Wigner
amplitude, and pair of second sheet poles of the S-matrix.
generality, because bound states can be added without difficulty. The

density matrix 3 should then be given by:
<Eaa-I W-(O)IE a a'-) = <Ea al wout(O)IEa a')
= F(Ea - ER)p(E)-l<all WRlla') (3.1)
where F(Ea - ER) is a Lorentzian energy distribution function with the
resonance width as the width of the energy distribution:
F(E _ )- ~ r /2 r 0
ER - 11: (ER _ E)2 + (r/2? lor :s,; E < 00.
(3.2)
Here we have normalized F(E - ER) in such a way that 4

F(E - ER) -> beE - E R) for r/2 -> o. (3.3)
As wet) goes into an observed out-state it will have to be the state W-(t)
and we shall use the basis vectors IE a-). We shall also normalize the reduced
matrix element to one:
(3.41
and discuss the normalization of W(O) = W R. We calculate
Tr W(O) = ~ <a-I W(O)la-) = ~ {'XlP(E) dE<E a-I W(O)IE a-)
=~ Ie dE F(E - ER)<all WRlla)
r roo 1
= 211: Jo dE (E - ER)Z + (r/2)Z
= ~ fOO dx _z_l_
11: x = - ER Zjr x +1
= ~ [~ + arctan ( E ~ R ) 1 (3.5)
where the new variable x is
(3.6)
The integration over E in the summation over the complete set of basis vectors
should extend over the spectrum of H, i.e., from 0 to 00. If the integration
over E would extend from - 00 to + 00 then the lower bound in the integral
over x would also be - 00 instead of - 2E R/r and we would have
Tr W R = Tr W(O) = 1. (3.5a)
3 The first line of (3.1) has been proven in Appendix XV.A.

4 See, e.g., Equation (II.8.12).
Equation (3.5a) can also be obtained from (3.5) in the limit of an infinitely
narrow width, i.e., if one takes instead of (XVIII.6.13) the limiting case
(3.7)
(because then arctan 2E R Ir -+ nI2). For sufficiently large 2E R lr, the exten-
sion of the integration over E to - 00 would make only a very small contri-
bution to (3.5). When we derive below the decaying state from the resonance
pole of the S-matrix, we shall see that the energy integration in (3.5) and
similar integrals will indeed extend from - 00 to + 00 and not from 0 to 00. 5
The expectation value of the energy operator in an energy eigenstate
W = 1 Eo)(Eo 1 is Tr(HIEo)(EoD = Eo. We shall now calculate the ex-
pectation value of H in the state W(O) = W R given by (3.1) which is to
describe a resonance with energy ER and width r:
(3.8)
In the limit (3.7) or if the integration over E extends to - 00, the first integral,
according to (3.5), goes to n, ~nd the second integral goes to zero, so that
(3.9)
This is the result we desire from comparison with the stationary energy
eigenstate.
Equation (3.1) gives only the diagonal matrix elements of the statistical
operator W R for a decaying system. To fix W R fully we also need to specify
its matrix elements between generalized energy eigenvectors of different
energy.
According to our discussion in Section II.9-in particular, according to
(11.9.1)- W(O) should in general be given by
W(O) = WR
= ~ fP(Ea)dEaP(E~)dE~f(E~ - ER)IE~il'->
aa
(3.10)
Hence
(3.11)
5 We should remind the reader that the Breit-Wigner amplitude for resonance scattering
must be an approximation, as remarked in Section XVIII.6. Therefore the choice, (3.2) with
the restriction on E may also be only an approximation and the more appropriate energy
distribution will be (3.2) with - 00 < E < + 00. As remarked, for sufficiently small 2r/ER
there will be only very little observable difference between these two.
so that by comparison with (3.1) we must have
2 -1 r 1
If(E - ER) I = pee) F(E - ER) = 2np(E) (E _ ER)2 + (r/2)2' (3.12)
Except for an arbitrary factor of modulus 1, there are only two possibilities
f£
from (3.12):
l
E-E - - (3.13)
f( R) - 2np ER + ir/2 - E
and
feE - ER)
rr 1
= ~2;p ER _ ir/2 - E' (3.14)
As we shall discuss in the following section, for a decaying state one has
to choose (3.14). Equation (3.13) will lead to a forming or "capture" state.
To simplify the notation we shall now specify the additional quantum
numbers a and then ignore them. As discussed in the preceding chapters,
the resonance R will have a definite value of angular momentum 1 = IR and
internal quantum numbers K = K R . Therefore a choice of angular-mo-
mentum eigenvectors
IEa-) = IEI1 3 K-) (3.15)
is convenient. The reduced matrix element of W R in this basis is
1
<K 11311 WRIIK' l' 1'3) = 8""R8uR8"'''R81'IR813/3 21 +1 (3.16)
for a completely unpolarized state. W R of (3.10) is then
WR= A+ I 13
fp(E) dEfp(E') dE'f(£' - ER)IE' IR 13 KR)
x <E IR 13 KR If*(E - E R), (3.17)

which we will write with (3.14) as
WR = ~ foo dE foo dE' 1

2n 0 0 ER - ir/2 - E'
1
x I£'-)<E-I ER + ir/2 - E' (3.18)
where we have defined for the sake of simplicity

IE-) = IE 113 K-)p1/2(E) (3.19)
and ignored the additional quantum numbers. The normalization and
summation for these new Dirac kets IE ±) is now fixed to be
<E'+ IE+) = 8(E' - E); I = foo dE (3.20)

E 0
instead of the general normalization given by (XIV.2.9) and (XIV.2.1O).

A proper, stationary energy eigenstate (pure if we ignore the additional

quantum numbers) of the observable H with eigenvalue ER is written as
(3.21)
where
(3.22)
and the 1ER) are proper eigenvectors and elements of the space of physical
states :6
(3.23)
In order to write the decaying state W R in an analogous way we define
the vector: 7
= ~ (~)1/2 JdE IE-) 1 . (3.24)

I 2n ER - ir/2 - E
This vector has the energy distribution
<E-lf D) = ~(~y/2 ER _ ir/2 _ E = ~ j<p=l)(E - ER)

= <fDIE-)*. (3.25)
The second equality in (3.25) follows if we treat 1 fD) and 1 E-) like well-
<
behaved vectors and I ) as their scalar product. Therefore we define the
conjugate vector <fDI by:
then the second equality in (3.25) is a consequence of this definition.

With (3.24) and (3.26), W R in (3.18) can be written as
W R = IfD)<fDI· (3.27)
The integration over E in (3.18), (3.24), (3.26) is to be taken from 0 to 00 if

we take the Breit - Wigner distribution (3.2) as the exact form. As we men-
tioned already, we will introduce only a minor error if we take the integration
from - 00 to + 00, extending the function feE - ER ) and therewith the
generalized vectors IE -) to negative values of energy. From the discussions
6 As it is important to distinguish between the proper and generalized eigenvectors, we shall

here use a separate notation IE) for a proper eigenvector and I E) for a generalized eigenvector
(cf. Section II.7).
7 Since we use for the IE -) the normalization (3.20), i.e. the convention peE) = I rather than
(XIV.2.9b), we have (0 make the same convention for the wave function; this is indicated by
the superscript (I' = I) oni
in Section XVIII.5 we are already familiar with the continuation to "un-

physical" values of the scattering energy. There we also learned that the
"unphysical" values of the energy cover the second sheet of a two-sheeted
Riemann surface cut along the positive real axis. The integral over the
spectrum of H runs along the upper rim of the cut of the physical sheet;
we will see below that when we extend it to negative values of E these wi11lie
on the second sheet.
The length of the vector liD) is given byB
D
(I II ) = 2n
D r If ,,- -
dE dE (E IE ) ER
1
+ irj2
1
_ E' ER - irj2 - E
if we use integration over the spectrum of H. It is equal to 1 if we integrate

from - 00 to + 00, which is sufficiently close to (3.28) for rj2E R ~ 1.
Equation (3.27) would be precisely the same as in the stationary case,
(3.21), if liD) was an eigenvector of H. We will show below that something
like (3.22) is indeed fulfilled for the liD), namely
HI/D) = (ER - i~)I/D) (3.29a)
and (if (/DI¢) = (¢I/ D)*):
(/DIH = (ER + i~) (/DI· (3.29b)
Equation (3.29) is not to be understood as a proper eigenvector equation

like (3.22) but as a generalized eigenvector equation valid if liD) is con-
sidered as a functional over roughly half of the space of physical states. 9
The time development of liD) is a formal consequence of (3.29):
(3.30)
As the time t for the decay process starts at t = 0, we need to consider
(3.30) only for t ~ o. It will turn out that (3.30) can be proven as a generalized
eigenvector equation precisely for these values of t.
8 J'"
-XR
dx-2- 1-
X +I
= ~ (!!.2 + arctan XR)
1t
= I - ~1t (~
XR
- ~ + ~ - ... ) for XR ~ 1.
3XR 5x R
9A self-adjoint operator cannot have complex proper eigenvalues but it can have generalized
eigenvalues which are complex.
The generalized eigenvector IjD) is according to (3.24) and (3.26) speci-

fied by the two real numbers ER , r or by the complex number
r
ZR = ER - i"2 (3.31)
which characterize a Breit - Wigner resonance. Following the usual conven-

tion of labelling the proper eigenvectors and the generalized eigenvector
with real eigenvalue (the Dirac kets) by their eigenvalue, it is appropriate to
also labelljD) by its eigenvalue and write
IjD) = IER - i~-) = IZR), (3.32a)
<jDI = (ER + i~-I = (Zrl· (3.32b)
With this notation (3.27) is written
(3.33)
This is the form for the statistical operator of a decaying state with resonance
parameters (ER' nthat is in perfect analogy to the form (3.21) for a stationary
state. W R has also, like a projection operator describing a pure state, the
property that
(3.34)
This follows immediately from (3.28) (with the integration extended from
- 00 to + 00). W R is however not exactly like AR and IjD) is not exactly
like IER ); this one notices immediately if one tries to calculate the length of
the vector HI fD) in the same manner as used in (3.28) (cf. Problem I).
For later reference we now give the expression for W R with the additional
quantum numbers (1 reinstated. Equation (3.10) is written in the form (3.27)
as
Here we have defined in analogy to (3.32) with (3.24):
IE R - rr
i~2 ' (1'-) =!i Vhi. JdE' p I/ 2(E')IE' (1'-)
ER -
1
ir/2 - E'
. (3.24a)
which is "normalized" in the sense of (3.28). The occurrence of pl/2(E) in

(3.24a) indicates that the IE' (1 - ) have the general" normalization" (XIV.2.9b).
For example, if the reduced matrix elements of W R are given by (3.16),
corresponding to a resonance with angular momentum lR and internal
quantum number K R , then W R becomes
(3.36)
XXI.4 Gamow Vectors and Resonance Poles of the S-Matrix 549
From the form (3.33) or (3.35), (3.36) of W R the exponential decay law
for the nondecay amplitude follows immediately. Using (1.4) and (3.30),
one obtains for t > 0:
Wet) = e-iHtiE R - i~-)(ER + i~-leiHt
= e- i(E R - I
ir!Z)ei(E R + ir(2) ER - i ~-) ( ER + i ~-I· (3.37)
Thus
(3.38)
For the nondecay probability (2.3a) we then obtain, with (3.34):
&>R(t) = Tr(WRW(t)) = Tr(WRWR)e- rt = e- rt Tr W R = e- rt . (3.39)
Equations (3.38) and (3.39) show that the state WR is an exponentially
decaying state. 10
Inserting (3.39) into (2.9) we obtain
&>(t) = 1 - e- rt (3.40)
from which it follows for the initial decay rate
#(t) It~O = r (3.41)
Comparing this with (2.5) and (2.7) we obtain
1 1
r =:.1. = r' (3.42)
Thus the lifetime of a resonance with resonance parameters (E R , r),

considered as an unstable physical system, is equal to the inverse width.
XXI.4 Gamow Vectors and Their Association to the

Resonance Poles of the S-Matrix
The present section will give a derivation of notions used in the previous
section and a mathematically more precise explanation of the statements in
that section. This section will make use of results obtained in Sections
XV.2 and XV.3.
In Sections XVII.5 and XVII.6 we related the poles on the negative real
axis in the physical sheet of the energy Riemann surface for the S-matrix,
10 As is well known to specialists. in the usual precise Hilbert space formulation one obtains
deviations from the exponential decay law [L. Fonda et al. (1978)]. Deviations for large values of
t follow from the condition that the spectrum of H be bounded from below which corresponds
to a finite lower limit in the integrals (3.5), (3.8), (3.24), (3.26) and which is not the case if one
integrates from - OCJ to + OCJ as we did to prove (3.34). Deviations from the exponential law for
small times t follow from the condition that the energy in the decaying state be finite, which
corresponds to the condition that the decaying state vector be in the domain of the Hilbert
space operator H (or even in <P, the domain of H) which is also not the case for the vector
I fD) as shown in Problem l.
the bound state poles, to eigenvectors of the energy operator H. These

eigenvectors with discrete eigenvalues describe according to Chapter XII
(pure) stationary states. For the stationary states we have, therefore, a dual
description by a pole of the S-matrix or by an eigenvector of H.
The continuous set of real energy values along the cut from 0 to 00 on the
physical sheet of the Riemann surface for the S-matrix correspond -accord-
ing to Section XVIII.5-to the physical values of the scattering energy. These
energy values are generalized eigenvalues of H (belonging to the continuous
spectrum). The generalized eigenvectors, or Dirac kets, belonging to these
eigenvalues describe idealized scattering states, i.e., their continuous super-
positions (e.g., (XV.2.7 +» or continuous mixtures (e.g., (XIV.5.27) with
(XIV.5.29b) and (XIV.5.36» describe pure scattering states ("wave packets")
or mixtures thereof. Thus for scattering states we have also a dual description,
either by the cut along the positive real axis of the Riemann surface for the
S-matrix or by generalized eigenvectors of H.
In the same way one can also describe the quasistationary states, defined
by the resonance poles of the S-matrix, as generalized eigenvectors of the
energy operator H. 11 These generalized eigenvectors, the Gamow vectors or
Gamow kets, will be functionals over roughly half the space of physical states
<1>, in distinction to the Dirac kets which are functionals over the whole of 
(cf. Section 1.7 or Section II.8). We shall now deduce the Gamow vectors
from the resonance poles of the S-matrix.
As in this section we are interested only in the principles we will first reduce
the problem to the essentials. For this purpose we choose the angular
momentum basis and ignore internal angular momenta (polarization):
la±) = IEa±) = IEIl 3 1]±)·
Further we restrict ourselves to just one channel, the elastic channel
I]= 1]' = I]A- In this case as a consequence of conservation of angular
momentum, (XVI.1.42), the S-matrix defined in (XV.3.lO') reduces to the
StCE) of (XVI.1.47). The relation that defines the S-matrix, (XV.3.10"), then
simplifies using (XVI. 1.43), (XVI. 1.45), and (XV.2.1O), to
IE 1131]+) = I PII'((EE» IE 1131]'-)<I]'IISllll])

~' P~
(4.1)
To simplify the notation to the essentials we will again use the definition
(3.19),
and write (4.1) as

(4.2)
Here we have also omitted the I from the S-matrix element.
11 Virtual states, defined by the poles on the negative real axis of the second sheet, can also be
described by generalized eigenvectors. We will not discuss this here but refer the interested
reader to A. Bohm, M. Gadella (l9R6).
We start our consideration from the S-matrix element (XV.3.2), (XV.3.14)

for the scattering of a pure physical state ¢in into a pure physical state l/JOUI
which we assume-for the sake of simplicity of notation-to have a well-
defined value for the quantum numbers a:
¢in = ¢in(a, 0) = ¢in(l13 11, 0),
l/JoUI = l/Joul(b,O) = l/JoUI(l1 3 11, 0). (4.3)
We then ignore these quantum numbers in the S-matrix element:
(l/JOUI(t), s¢in(t)) = (n-l/JOUI(t), n+ ¢in(t)) = (l/J-(t), ¢+(t»)
= (l/JOU\ S¢in) = (l/J-(O), ¢+(O») = (l/J-, ¢+). (4.4)
We now expand l/JOUI and ¢in according to (XV.2.6±) with respect to the
basis system of eigenvectors of K (XV.l.l4):
¢in = L fp~(E)dE IE 11 3 11)<E 113 11 I¢in) = fdE IE)<EI¢in),

113~
l/JOUI = L fp~(E)dE IE 11 3 11)<E 113111l/JoU1 ) = fdE IE) <Ell/IoUI), (4.5)

113~
where we have used for the unperturbed basis vectors the simplifying notation
The integration in (4.5) extends over the spectrum of K which we have

assumed to be identical with the continuous spectrum of H, Spect.H,
which we choose (without restricting generality) to be from 0 to + 00.
It is important to mention that ¢ + represents the pure physical state
which developed from the prepared in-state ¢in; it is therefore a well-behaved
vector. Similarly l/J- represents a pure physical state that develops into the
measured out-state l/IOUI; it is therefore also a well-behaved vector
l/J-- E <D c ;Yf c <DX. And so are ¢in and l/JOUI. The energy wave functions
<EI¢in) = <£+ I¢+) (given by the incident beam),

<EIl/JoUI) = <£-Il/J-) (measured by the detector)
are therefore well-behaved functions of energy. The equality in (4.6±) is a
special case of (A.2 ±) in the Appendix to Chapter XV.
Inserting (4.5) into (4.4) and making use of (4.6±), the S-matrix element
(4.4) can be written (omitting all inessential quantum numbers):
(l/J-, ¢+) = lt~ l.t. ff dE' p~.(E') dE p~(E)

x <l/JOUI 1£' l' 1~ 11') <£' l' 1~ 11' 1S 1£ 113 1'/) <£ 113 111 ¢in)
= r
J Spec!. H
d£<l/J-IE+ic-)S(£+iO)<£+iE+I¢+). (4.7)
Here we have used (XVI.I.43), (XVI.I.45), (XVl.I.4 7), and the assumption that
4;in and t/J0ut have the same value for 113 1] and avoided the irrelevant compli-
cation in the notation by using (3.19). The integration in (4.7) is along the
spectrum of H which is along the upper rim of the cut of the physical sheet
of the S-matrix. We have indicated this by writing E + iO as the argument
of the S-matrix. Integration along the upper rim of the cut in the physical
sheet is, according to the property of the S-matrix described in Section
XVIII.5, the same as integrating along the lower rim of the cut in the second
sheet because SeE + iO) = SeEn - iO), where the subscript II refers to the
second sheet.
We want to deform the contour of integration of (4.7) into the lower half
of the second sheet of the Riemann surface for the S-matrix, where-
according to Section XVIII.5-the resonance poles are located. As in this
section we are only interested in the principle that establishes the connection
between resonance poles and Gamow vectors, we will make the simplest
possible assumption about our resonance model. We assume that our
physical system is such that the S-matrix has no bound state poles and no
singularities other than just one pair of resonance poles. 1 2 This is depicted in
Figure 4.1. Figure 4.l(a) shows the upper half of the first sheet and the lower
half of the second sheet with the resonance pole at Z R = E R - ir/2. Figure
4.1(b), showing the upper half of the second sheet with the resonance pole
at Z~ = ER + i(r/2) and the lower half of the first sheet, will be used in the
latter part of this section.
The contour of integration of (4.7), which was originally along Spect. H,
has been deformed into the contour ~ _ and a circle around the pole. This
contour deformation is possible if the well-behaved wave functions
<,P-IE-> and <E+14;+>, which were initially only defined on the spectrum
of H, are boundary values of analytic functions in the lower half of the second
sheet. For further deformations of the contour we also want to assume that
these analytic functions vanish sufficiently fast at the lower infinite semi-
circle,u This is a condition on the set of well-behaved vectors t/J -; not
all vectors representing physical states fulfill this condition. Thus the set
of vectors t/J - for which the contour of integration in (4.7) can be deformed
into the contour shown in Figure 4.1(a) cannot be the set of all pure physical
states, i.e., cannot be the set of all vectors in <1>. But if we consider a decay
process only, i.e., the later half of a resonance scattering process as depicted
by Figure 1.2, then the {t/J -} need not represent all the pure physical states
but only roughly half of them, namely only those which start developing
at t = 0 and go for t ...... x into the observed out-states t/J0ut. The remaining
12 If there are more than one pair of resonance poles one obtains for each pair a pair of
Gamow vectors and if there should be other singularities the background term introduced
below will have a different form, but otherwise the arguments carry through in exactly the
same way. If bound states exist one must carefully distinguish between Yf and Jf;cat (cf. Chapter
XV). but otherwise our arguments will not be affected.
13 The S-matrix in the second sheet is polynomially bounded at infinity for many interaction
potentials. Assuming this to be the case also for our model, we need wave functions which
vanish sufficiently fast so that the integrals over the infinite semicircle are zero.
(a)
E (first sheet)
Sp H
E(second sheet)
(b)
Figure 4.1 Deformation of the path of integration into the second sheet of the energy
plane. (a) for the decaying state; (b) for the growing state. [Reprinted from A. Bohm,
J. Math. Phys. 22, 2817 (1981).J Sp R denotes the spectrum of the (closure ot) the
Hamiltonian.
half of the pure physical states is the set {If; +} of vectors If; + which come from
prepared in-states If;in for t -+ - CIJ and stop developing at t = O. This set
will be needed when we consider the contour deformation into the upper
half of the second sheet in the discussion of the formation process depicted
in Figure 1.2(b).
The physical separation of the set of states has its mathematical counter-
part in the division of the space of physical states <l> c £ c <l> x • In order to
show this we have to introduce some more mathematics:
[The Hilbert space of square integrable functions on the real line H

(a possible realization of the space £) is the direct sum
H = H~ Er> H~ (4.8)
of the space of Hardy class functions from above H~ and the space
of Hardy class functions from below H~.
A complex function G(E) on the real line is a Hardy class
function 14 from above (below) if
(1) G(E) is the boundary value of a function G( OJ) of the complex
variable (complex energy) OJ = E + i1J that is analytic in the
half plane I] > 0 (I] < 0)
14 P. L. Duren, Theory of HP-Spaces, Academic Press, New York, 1970.

(2) f~oo 1 G(E + il]W dE < k < 00
for aliI] with ° < I] < Xl ( - Xl < '7 < 0).

The function G(E) is called an H'+-class function if the power 2 in
the integral is replaced by the power p = 1,2, 3, ....
Hardy class functions have the following important properties:
(A)15 If G(E) E' H~ or G(E) E' H~ then G(E) (on the real axis) is
uniquely determined by its values on the positive real axis.
(B)16 If G(E) E' H'+, 1 0,
= 1m w < 0.
If G(E) E' H~, 1 0,
-2ni -00 E_w dE = G(w) forl]=lmw<O.
This is often referred to as the Titchmarsh Theorem.
If G ± (E) E H~, p = 1, 2; then the Fourier transform
f
(C)17
G+(t) =-~- + 00 e-itEG +(E) dE

- .jbc -00 -
has the property

for t ~ 0.G+(t) = °
This is called the Paley-Wiener Theorem.
(D) If G(E) E' H'+ then its complex conjugate G*(E) belongs
to ~ and vice versa.
We will call a well-behaved vector 1jJ- E' <1> a Hardy class vector
from above, and write 1jJ- E' <1>+, if its wave function <E-IIjJ-) E' H +
(precisely if the values of <E-IIjJ-) on the spectrum determine
by property A a Hardy class function from above). Then
<IjJ -I E - ) E' H_. Similarly we call IjJ + E' <1> a Hardy class function
from below, and write 1jJ+ E<1>_, if <E+ 11jJ+) E H _.
In this way the space of physical states separates into two non-
intersecting spaces <1>+ and <1>_. We mention without further dis-
cussion that they have a nuclear locally convex topology and each
of them is dense in ~ where Yf is the space realized by the square-
integrable functions on the spectrum, i.e., the positive real line, not
by H. Thus we can construct two rigged Hilbert spaces 18
<1>+ c £' c ~ and <1>_ c £' c cD~, (4.9)
where cDt±) is the space of continuous anti linear functionals on cD( ± ).]
15 C. van Winter, J. Math. Anal. 47, 633 (1974); Trans. Am. Math. Soc. 162,103 (1971).
16 P. L. Duren, ibid., Theorem 11.8.
17 P. L. Duren, ibid., Theorem 11.9.
18 For details see A. Bohm, M. Gadel1a (1986).

After this mathematical interlude we can continue with the discussion of

the contour deformation in the integral (4.7). Now we admit to (4.7) only
those tjJ- which can represent the state of decay products tjJ0ut for t -+ 00
that start to appear only at t = 0 (for t < 0 such decay products have not
yet come into existence). This we then translate-on the basis of the discus-
sion in the mathematical insert-into the mathematical statement
tjJ- E <1>+ (4.l0a)
which means
(4. lOb)
and
En<E-ltjJ-)EH~, En<tjJ-IE-)EH~ forevery n=1,2,3, ....
(4.lOc)
(4.10c) is fulfilled because <E-ltjJ-) is well-behaved (i.e., tjJ- E <1».
After the contour deformations have been performed, assuming that
<
<tjJ - 1w -) and w + 1(V) have the mathematical properties required for it,
(4.7) goes over into
(tjJ-, ¢+) = L- dw<tjJ-lw-)Su(w)<w+I¢+)
(4.11 )
Here SIl(W) = Sewn) denotes the value ofthe S-matrix in the second sheet and
d _ I is the residue of S" (w) in the second sheet at Z R :
SII(W) = d_ 1 + dO + dl(W - ZR) + .... (4.12)

W - ZR
The second integral in (4.11), the integral around the pole at ZR will be
considered separately below; it is this integral which will give the Gamow
vector. We will first consider the integral along CfJ_ which we call the back-
ground. As the integrand does not have any further singularities in the
lower half of the second sheet, we can deform the contour CfJ _ further and
write the background as an integral from 0 to - 00 plus an integral over the
infinite lower semicircle in the second sheet, which however vanishes (because
SIl(W) does not grow faster than a polynomial when Iwl-+ 00). Thus for the
background we obtain:
background = {- oon dEli <tjJ- 1 Eii )S(Eu - if) <Eii (V).
1 (4.13)
Here we have indicated by Ell that the integral is running along the negative
real axis in the second sheet and by - 00 n that it goes to infinity on the
second sheet.
We now consider the pole term. Using the Cauchy integral formula (A.1a)
of the Appendix to Chapter XVII we obtain for the second integral in (4.11):
(4.14)
Adding (4.13) and (4.14) we can write (4.7) in the form
(I/I-,¢+) = fo-a:JndEu(I/I-IEu- jf+)(Eu - jc+I¢+)
+ (1/I-IZ;)(2nr)(Z~ I¢+)· (4.1Sa)
Here we have used (4.2) extended to "unphysical" values of E:
(4.2a)
We have also made use of the fact that as a consequence of the unitarity the
residue of the S-matrix is related to the width by
(4.16)
which can be immediately seen by comparing (4.12) with the representation

(XVIII.6.23') of the S-matrix. 1/1- was an arbitrary vector except that it had
to be an element of <1>+. We can omit this arbitrary vector in (4.1Sa) and
write (4.ISa) as
¢+ = 1- dEIIIEIt>(E~I¢+)
00
+ IZ;)(2nr)(Z;I¢+) on <1>+. (4.1Sb)
This looks very much like the generalized basis vector expansion (II. 7.16),
(XV.2.7 +), or (1.4.4g) except that for those basis vector expansions one
could have taken the scalar product with a 1/1- which could have been from
the whole space <1>, whereas in (4.1Sb) one can only take the scalar product
with 1/1 - E <1> +. In other words, the generalized basis vectors IEIt> and
IZ;) are only defined as functionals over <1>+, whereas the Dirac kets
IE +) with E E Spect. H are defined over all of <1>.
The pole term integral in (4.14) does not yet look as ifit has anything to do
with a resonance. A resonance has a Breit - Wigner energy distribution, so
we will perform further contour deformations to bring the integral in (4.14)
into a form exhibiting the typical Breit - Wigner amplitude. The circular
contour in (4.14) is deformed into a straight line from - 00 to + 00 right
below the real axis plus an infinite semicircle. If the integral over the infinite
semicircle vanishes then (4.14) takes the form
(4.17)
This integral contains the Breit-Wigner distribution 1/(EII - ZR) but it

extends over the entire real axis.
Equation (4.17) is the same statement as Theorem B (the Titchmarsh
theorem) for the function G_(En) = (1/I-IEi.)(E.;I¢+)Lt. Thus in order
that the Breit - Wigner integral emerges from the pole term the function
G _(En) must be an element of H ~ .
If we use the theorem B for <1/1- IEii) E H~ then
1 f+oo 1
- -2. dE n <1/I-I Eii) E _ Z = <1/I-IZ;). (4.18)
1tl -00 II R
Omitting in (4.18) the arbitrary 1/1- E+ we get a definition of a vector

IZ;) which has meaning only as a functional over +, which means that
one can only take the" scalar products" of it with vectors which are E + :
IZ;) = - -2.
1
nl
f-
+ 0011
0011
1
II R
( 1 ) 1/2
dEli lEi,) E _ Z == 2 r
n
IjD)
_
= ( -1-
2nr
)1/21 ER - 1-
. r-)
2
. (4.19)
This vector is-except for a "normalization" factor-identical to the

vector IjD) of (3.24), as indicated by the second equality in (4.19). This
vector we call, because of its time development derived below, the Gamow
vector or Gamow keto We now also understand why jD in (3.24) is to be
defined by an integral over the Breit - Wigner amplitude from - 00 to + 00
and not from 0 to 00 as one would have expected from the spectrum of H:
The theoretically correct Breit - Wigner amplitude, i.e., the one connected
with the resonance pole of the S-matrix, has its support on the whole real
axis, more precisely from - 00 to 0 on the second sheet and from 0 to 00
on either the second sheet immediately below the cut or on the first sheet
immediately above the cut. The" normalization" integral in (3.28) should,
therefore, extend from - 00 to + 00 rather than from - 2E R/r to 00 so that
<jD IjD) = t.I 9 However, as already mentioned, for practical purposes the
distinction is imperceivable if r/ER ~ 1.
The complex conjugate of (4.18) gives
(4.20)
Using for the left-hand side Theorem B this leads to
(4.21)
19 To avoid misunderstanding concerning the mathematics it must be emphasized that the

Breit-Wigner amplitude (E- I fD) is an element of H in which the scalar product is defined
by integration from - 00 to + 00. If D) is not an element of Jf, but an element of<ll~ (functionals
over <II +). .Yt is realized by the space of square integrable functions on the positive real line
U([R+) (the spectrum of H). not by H = U([R). The physical scalar product is given by the
integral over the spectrum or also-according to (4.15b)-by the integral from 0 to -ooplus
the integral from - 00 to + 00 which differs from the scalar product in H by the integral from
o to - 00. Therefore vectors in H become generalized vectors for .rf.
Again omitting the arbitrary lIP-) we obtain
<Zrl = -2'
1 fro 1
dEII<EiI IE _ Z*
nl - 00 II R
== (_1 )1/2 <fDI == (_1 )1/2(ER + i!::-I (4.22)

2nr 2nr 2
which is the same as (3.26).
Using the "normalized" vectors and (4.16) the pole term (4.14) can be
written
(4.23)
and the generalized basis vector expansion (4.15b) has the form
(V = fooodEn'EIt><Elt'(V) + IER - i~-)( ER - i~+I(V) (4.15c)
which is an equality between continuous functionals in ~. If instead of one

resonance pole one has n resonances Rb R 2 , ••. , Rn then the last term will
be replaced by a sum of n such terms. Thus in the new basis vector expansion
(4.l5a)-(4.15c) the resonances appear in very much the same way as the
bound states in the ordinary spectral decomposition.
We shall now derive (3.29). To this end we consider the vector Htf;-. By
assumption (well-behavedness) this is again EO <1>+. We can then repeat the
whole above procedure starting with (4.7) and replacing everywhere tf;-
by Htf;- on the left-hand side and <tf;-IE-) by E<tf;-IE-) under the
integral. When we reach (4.18) it will now read
(4.24)
Since E <tf; - IE - ) is also a Hardy class function from below we can apply the
Titchmarsh theorem with G(En) = E n<tf; -I Eii) to the right-hand side and
obtain
(4.25)
This is the precise formulation of the statement that IZ R) is a generalized
eigenvector of Hover <1>+ with eigenvalue ZR' which was written formally as
(3.29a). Equation (3.29b) is then immediately obtained by taking the com-
plex conjugate of (4.24) and using the Titchmarsh theorem for the Hardy
class function from above G(E) = E<E-Itf;-).
Equation (3.30) may appear to be an immediate consequence of (3.29), but
then one would wonder about the restriction t ~ O. It is indeed not straight-
forward. I fD) = IZ;) is an element of ~, the space of continuous anti linear
functionals over <1>+. Thus the unitary time development operator U(t) =
e- iHt in the Hilbert space has to be extended to a time development operator
in ~. This can only be done if U*(t) = eiHt leaves the space + invariant,
i.e., if and only if eiHttf; - E + for every tf; - E + .
We will now apply the above procedure starting with (4.7) to the vector
eiHttf;- instead of tf;-. Under the integral on the right-hand side of (4.7) we
then have e-iEt(tf;-IE-) in place of (tf;-IE-). Everything in the above
procedure can be repeated as long as e-iEt(tf;-IE-) is a Hardy class function
from below. But if G(E) E H ~ then e - iEtG(E) E H~ iff t ~ O. Therefore only
for t ~ 0 can one repeat the above procedure and arrive in place of (4.18) at
( e''H t'/'-IZ-)
R
- I fdE'IIE
= - 2' e-' IIt('/'-IE-) _1__
II E11- Z' for t ~ 0 only.
0/
m 0/
R
(4.26)
To the right-hand side one can again apply the Titchmarsh theorem and
obtain
(eiHttf;-IZi) = e-iZRt(tf;-1 Zi) for t ~ 0 and every tf;- E <1>+.
(4.27)
Equation (4.27) is the precise form of (3.30).
If eiHt is a continuous operator from + to + one can define the conjugate
operator in <1>: by the usual definition of the conjugate operator in the rigged
Hilbert space: 20
(4.28)
This operator, defined only for t ~ 0 is what is meant by the operator in
(3.30):
(ei{!t) x == "e- iHt ", t ~ o. (4.29)
Here the subscript + indicates that it is defined in <1>: only, where it is the
extension of the Hilbert space operator e - iHt.
We have herewith established all the results that have been used and
justified only heuristically in Section XXI.3. We have given more precise
definitions and statements of Section XXI.3's new entities and their proper-
ties. And we have shown that the exponentially decaying state, the Gamow
vectors (4.19) or (3.24), come from the resonance pole of the S-matrix in the
lower half plane.
To complete our discussions we shall now very briefly demonstrate that
the resonance poles above the real axis lead to exponentially growing states.
We start from the complex conjugate of (4.7) where we use the symmetry
relation of the S-matrix (XVIII.S.lO). Then we obtain
(4) +, tf;-) = f
Specl. H
dE (4) + IE+)S*(E - iO)(E-Itf;-)· (4.7')
If the integration in (4.7) was along the upper rim of the cut the integration
is now along the lower rim of the cut in the physical sheet, which is indicated
by writing E - iO as the argument of S*. We use the same symbols 4>+ and
tf; - as above but these 4> + and tf; - have now an entirely different physical in-
terpretation and therefore different mathematical properties than those
20 For example, A. Bohm (1978), Chapter III.

entities denoted by the same symbol above. Indeed, what we consider now
is the earlier half of the resonance scattering process in which one starts in
the remote past, t ~ - 00, with an in-state ¢in that becomes ¢ + developing
with the Hamiltonian H into a state ¢+ at t = O. Thus whereas in (4.7) 1/1-
described the controlled state, i.e., the state determined by the experimental
arrangement for the decay products l/1out, in (4.7') it is the vector ¢ + which is
the controlled state. ¢+ in (4.7') is controlled by the preparation apparatus
for ¢in and 1/1 - in (4.7') is the state that results as the consequence of the
dynamics.
For these controlled states ¢ +, which will have to be gone by the time
t = 0, and which therefore constitute roughly half of the space of physical
states which we call <1>_, we postulate that the wave functions <E+ 1 ¢ +) are
Hardy class functions from below (because then S<E+ I¢+(t) dE = 0 for
<
t > 0 by the Paley-Wiener theorem). Thus ¢ + 1E +) in (4.7') are E H t
and we may deform the contour of integration in (4.7') through the cut into
the upper half plane of the second sheet as shown in Figure 4.1(b). (4.7') then
goes into:
(¢+, 1/1-) = f dw <¢+lw+)Sn(w)<w -11/1-)
'e+
(4.11')
where L 1 is the residue of Sew) in the second sheet at Z~ :
(4.12')
and «J + is the path above the pole at Z~ and 0 is the circular path around Z~
as shown in Figure 4.1(b). Again assuming sufficiently nice properties of
<¢ + 1w +), <w - 11/1 -) and S(w), the first integral, the background, can be
written
background = {-<XlII dEn <¢+IE DS(En)<Eilll/1-).

1 (4.13')
The pole term is evaluated using the Cauchy integral formula (XVII.A.1a)
and we obtain
(4.14')
The integral in (4.7') can therefore be written as
(¢+, 1/1-) = {-<XlII dEn <¢+ IEI~)<Eilll/1-)

+ <¢+ IZr)2nr<Zr 11/1-), (4.15'a)
where we have used an extension of (4.2) to the second sheet,
(4.2')
and the consequence of the unitarity of the S-matrix 21
(4.16')
The integral in the pole term (4.11') is now transformed into a Breit - Wigner
integral by the contour deformation using the fact that
«VIE,t><Ei,l,r>L 1 E H~:
(4.17')
<(V IE+ >is now E H~, i.e., cfJ + E <1> _. The Titchmarsh theorem applied to
this function then gives
+ *+ > -_ 2----:
<cfJ IZR
1 f+oo + + 1
dEn <cfJ lEn> E _ Z*' (4.18')
m -00 n R
Writing this without the arbitrary vector cfJ + E <1> _, we obtain
IZR
*+ >-_ 2----:
1 f+oo +
dEn lEn >E
1
- Z*
m -00 n R
(4.19')
r
The vector I i >= IE R + if/2 +), G standing for growing or forming,
defined by (4.19') is the Gamow vector associated with the pole of the S-
matrix in the upper half of the second sheet. It has meaning only as a func-
tional over <1> _. Thus there are two Gamow kets associated with each
resonance, the exponentially decaying functional over <1>+, (4.19), and the
functional (4.19') which is exponentially growing, as we shall derive below.
The complex conjugate of (4.18') leads to
<Zil = - -.
1 f+ 00 1
dE<E+I--
2nl -00 E - ZR
(4.22')
21 Note that in (4.16) an arbitrary phase factor has been fixed (incorporated in the definition
of the kets which also contain an arbitrary phase). The phase in (4.16') is then no longer arbitrary
but is given by time-reversal invariance.
In place of (4.l5c) we obtain now from (4.l5'a) for the generalized basis
vector expansion over <1>_
t/J - = L-00 dElI IEll> <E iIi t/J - > + IE R + i~ +) (E R -I

+ i ~ t/J - )
(4.15'c)
which is an equality between continuous functionals in ~. (Note that the
symbols ¢ + and t/J - have different meanings in the primed and unprimed
equations. )
The proof that IZr >is a generalized eigenvector of H with eigenvalue
Z'k = ER + if'/2 is a straightforward application of the Titchmarsh theorem
to Ell <¢ + I E,t> E H~. In the same way as for (4.25) one shows
<H¢+IZ'k+> = Z'k <¢+IZr> for every ¢+ E_ (4.25')
which one also writes formally as
Hlf G> = (ER + i~}fG>, (4.25'a)
<fGIH = (ER - i~}fG>. (4.25'b)
The time development of I fG> is now obtained in complete analogy to

(4.27). We have to define a time-development operator in ~ as an extension
of the Hilbert space time development operator e- iH ,. This operator can only
be defined as the conjugate operator to the operator eiHt in _ if for every
¢ + E _ also eil/' ¢ + E _. This means that it can be defined only if with
<¢+IE+>EH~ also <eiH'¢+IE+> = e-iEt<¢+IE+>EH~. This is the case
only if t ~ O. Therefore one obtains for ¢ + E _ from
(4.26')
by application of the Titchmarsh theorem to right-hand side:

<eiHt ¢ + IZ~ +> = eiZRt <¢ + IZ~ +> only for t ::;; 0 and for every ¢ + E <1>_ .
(4.27')
Thus one can define the time development operator in <1>: as the conjugate
of the operator eiHt in <1>_ only for t ~ 0:
t ::;; O. (4.29')
This is the extension of the unitary time-development operator e - iHt in
'y{ to the space <1>." . Formally one writes (4.27') as
(4.27'a)
keeping in mind that one can take the" scalar product" of this equation
only with elements of <1>_. One also writes
"e-iHt"lfG><fGI"eill,,' = er'lfG><fGI for t ~ o only. (4.27'b)
XXI.5 The Golden Rule 563
This is to be compared with (3.38) for the exponentially decaying state. Thus
we see that the Gamow vector I fG), which derives from the resonance pole
above the real axis, represents an exponentially growing state.
XXI.5 The Golden Rule
The formula that expresses the initial decay rate in terms of the T-matrix
or the matrix elements of the interaction Hamiltonian V is called the Golden
Rule. The decay rate is-according to (XIY.2.3)-the time derivative of the
decay probability (2.8), which is the transition probability from the decaying
state Wet) = e-iHtl fD)<fDle iHt into the states of the decay products de-
scribed by the subspace AYf. Though one could use the results of Section
XIV.2 and derive immediately the decay rate (XIV.2.6) or (XIV.2.l8), it is
instructive to derive first the expression for the decay probability
&(t) = Tr(A Wet)) (5.1)
and then obtain from it the conventional form of the Golden Rule for the
initial decay rate :?J>(t = 0) by differentiation.
In order not to obscure this simple derivation by unimportant details we
will first ignore all the additional quantum numbers a(= l/3 K) and only
afterwards take into consideration that the state W R is a mixture of I fD) =
IER - ir/2, a-) as expressed by (3.35).
AYf describes the final decay products after the interaction has ceased.
Therefore-as in Section XIV.2-
A = L Ib)<bl, where Ib) = IE b , 6)
b
are eigenvectors of the free Hamiltonian K. Using these basis vectors for the
calculation of the trace in (5.1) we obtain
&(t) = L <ble-iH'lfD)<fDleiHtlb). (5.2)
b
The sum in (5.2) goes over all values of b = (Eb' 6) if A is the projection
operator on the space of physical states of all decay products into which the
decaying state I fD) <fDI can decay; if only a part of the decay products are
observed, i.e., if only a partial decay rate is measured, then A is the projector
on the subspace of these observed decay products and the sum in (5.2) is only
over those basis vectors which span this subspace.
The vectors Ib) in (5.2) are replaced by the Vlb).and the eigenvectors of
H, I b -), using the solution of the Lippmann-Schwinger equation (XV.1.22).
One obtains
<JDle-iH'lb) = <fDle-iHtlb-) _ <fDle-iHtVlb) 1 . . ,
Eb - (ER + Ir/2) - IE
(5.3)
where the property (3.29b) of the decaying state vector <f v I = (ER + ir/2 - I
was used.
Inserting (5.3) into (5.2) gives:
PJ(t) = {Xl dEbe-iEbl<E; IZi)(Zr IE;>eiEb'

- L (Z~ -I eiH1lb- >--
Eb - (ER
1.
+ Ir/2) - If
.
+ ~ <bl Ve-iHIIZi)(Z~-leiHIVlb> (Eb _ ER)~ + (r/2)2· (5.4)
In the first term we have replaced Lb

by the integral over the spectrum of H
which is identical with the spectrum of K. The second and the third term in
(5.4) are the complex conjugates of each other. We now use the property
(3.30) = (4.27) of the decaying Gamow vectors; (5.4) then becomes
PJ(t) = {Xl dEb <E; IZi)(Zr IE;>
Here c.c. denotes the complex conjugate of the preceding second term,
which we denote by II. This term is now elaborated further by using again
the solution of the Lippmann-Schwinger equation (XV.1.22-) for Ib- >:
II = _e- rr t «Zr Ib><bl VIZi) Eb _ (ER _ ir/2) + if1
1
+ (Eb - ER)2 + (r/2? (Zr I Vlb><bl VIZi»·
The first term in II is zero. The reason for this is analogous to the reason
that led to II = 0 following equation (XIV.2.17) or preceding equation
(XV.A.4):22 (Zr I describes the decaying state, Ib> spans the space of the
decay products when the interaction V performing the transition is switched
off; thus (Zr Ib> is the probability amplitude for no transition, whereas
<bIVIZi) is the probability amplitude for the transition. The second term
in II is real, therefore c.c. in (5.5) is identical to II. This is also identical
(except for the opposite sign) to the fourth term in (5.5). The first term in
22 Instead of using (5.1) as the starting point of our derivation we could have started with
(XIV.2.18) for the decay rate, in which this term is already gone. This would however have
made the calculation slightly less transparent, cf. Problem 3.
XXI.5 The Golden Rule 565
(5.5) is obtained from (3.28) to be equal to unity, except for a small correction
r/2nE R + .... Thus we obtain for the decay probability:
&(t) = (1 - ~
2nE R
+ ... )
- e- n t <blVlfD)<fDlVlb) (Eb _ ER); + (r/2)2' (5.6)
This formula for the decay probability of a quasistationary state with energy
energy ER and width r does not contain any approximation. It has the form
expected from (3.40) for r/2E R --> 0 with
2ne- n (r/2n)
&R(t) = -r-~ l<blVlf D)1 2 (Eb _ ER)2 + (r/2)2 . (5.7)
According to (3.39), &R(t) at t = 0 should be equal to 1 (and &(t) --> 0).

As (cf. Equation (11.8.12)):
r/2n
(Eb _ ER)2 + (r/2)2 --> J(Eb - ER) for r/2ER --> 0 (5.8)
we obtain from (5.7):
r = 2n L l<blVlf D)1 2 J(Eb - ER)' (5.9)

b
For the decay rate we obtain from (5.6) by differentiation with respect to
time:
. _ -n" 2 r/2n
2nfl<blVlf)1 (Eb _ ER)2 + (r/2)2
&(t) - e
D
-fr JOO D 2 r/2n (5.10)

= e 2n 0 dEb I<Ebl Vlf )1 (Eb _ ER)2 + (r/2)2'
This formula for the decay rate contains no approximation. The factor
r/2n
(5.11)
b(Eb) = (Eb _ ER)2 + (r/2)2
is called the natural line width. It can be observed if the detector has a very
good energy resolution fl.Eb ~ r so that one can measure the decay rate
per unit energy:
d#(t) -n D 2 r/2n
dEb =e 2nl<EblVlf)1 (E b -E R)2+(r/2)2' (5.12)
In the limit of narrow width r/2E R --> 0 the initial decay rate, (5.10) for t = 0,
becomes
#(0)=2n L l<blVlf D)1 2 J(Eb - ER)' (5.13)
b
Comparing this with (5.9) we see, as already observed before (3.14) that
the initial decay rate is equal to the width.
The formula (5.13) is the conventional form of the Golden Rule. 23 As

(5.6) to (5.12) are different forms of one and the same result we will refer to
all these formulas as the Golden Rule.
Though (5.6), (5.10) and (5.12) do not entail any approximation, they
contain some quantities which are not easily accessible and can usually only
be obtained by approximation methods-like the generalized eigenvalue
of the exact Hamiltonian, ER - if/2, and its eigenvector IfD). For narrow
resonances, like the excited states of atoms and molecules and in any other
case when the interaction Hamiltonian V describes a weak interaction, r is
orders of magnitude smaller than ER and I fD) can be replaced by the eigen-
vector of K which is stationary with respect to the time development e- iKt
(Born approximation).
Now that the fundamental results have been derived for the simplest
model, we will reincorporate the additional quantum numbers ain order to
apply the Golden Rule to realistic cases. This is done by replacing I fD) fDI <
in the above formulas with W R of (3.35). The result for the decay rate is
&>(0) = 2nLLb(Eb - ER)<bIVIER,a')(ER,alVlb)<a'IIWRlla) (5.14)
baa'
where we have written IE R, a) for the decaying state vector IER - ir;2, a-).
For the familiar case of a decaying state with angular momentum IR and
internal quantum numbers K R [Equation (3.16)J, this goes over into
.
9(t = 0) = 2n I1L L " " .(j(Eb - ER)<bl VIER, IR' /3' KR)I 2 . (5.15)
(2 R + 1) 13 b
(5.14) is the general form of the Golden Rule for the initial decay rate. If the
summation in b extends over all values of the quantum numbers b = (Eb' b)
then P is called the total decay rate.
Equations (5.6)-(5.15) are formulas used for the calculation of decay
rates in all branches of physics. Usually the interaction is weak, so that the
Born approximation can be used for the transition matrix element
<bl VIER, a). This means IE R, a) is replaced by the eigenvector of the free
Hamiltonian
K=H-V. (5.16)
The I b) are already eigenvectors of K [cf. (XIV.2.14)]. So the program for
the calculation of the initial decay rate is the following: One finds the eigen-
states and generalized eigenvectors of the free-energy operator. These are
the energy eigenvectors in the approximation in which the decaying system
is considered stable. For instance, in the decay (1.3a), IE R, lR' /3, KR) will be
the proper eigenvector of the energy operator of the atom when the interac-
tion with the radiation is ignored and all energy levels are stable. Ib) is the
direct product
23 This formula of Dirac has given so much praiseworthy service in many areas of quantum
physics that Fermi called it a Golden Rule (the "Golden Rule No.2").
XXI.6 Partial Decay Rates 567
where the first factor is the ground-state energy eigenvector of the atom,
and Iy> is a basis vector of the one-photon space. Then one has to conjecture
the interaction Hamiltonian V (or the transition operator T if the Born
approximation is not satisfactory). For the case of the radiative decays of
atoms this can be done to a certain extent by using the correspondence with
a classical system; in other situations, e.g., for the decay of elementary
particles, one has only very few general principles to limit the possibilities
for this conjecture. After V and the solution of the free problem are known,
one has to calculate the matrix elements of V between the free eigenvectors
and insert it into (5.14) or (5.15). In this way the transition is considered as a
transition from one "stable" state into another, caused by the interaction
Hamiltonian Vor by the transition operator T.
XXI.6 Partial Decay Rates
Even more important in practical calculations than the total decay rate are
the partial decay rates. In order to obtain a partial decay rate we have to
specify the basis vectors I b) in (5.14) or (5.15). For the sake of definiteness we
shall give our discussions first in terms of the angular-momentum basis.
One can choose
Ib) = IE b , b) = IEb,jh 1]) (6.1)
if the angular momentum commutes with the internal observables:
(6.1a)
The internal observables I]0P will in general not agree with the internal ob-
servables /Cop of which the resonance is an eigenstate with eigenvalue K R . We
then write (5.15):
r=2nIfp~(E)dEi5(E-ER)II 11 I<Ejj31]1VIE R IR I3 /CR)1 2 .

~ jh 13 2R + 1
(6.2)
This gives the rate for the decay of R into decay products with any of the
internal quantum numbers 1]. If the detector detects only decay products
with one particular value for 1], then one measures the quantity
r(R ~ 1]) = 2n f p~(E) dE beE - E R)

1
x I I
jh 13
21
R +
II<Ejj31]1VIERIRI3KRW, (6.3)
r(R ~ 1]) is called the partial decay rate for the decay of R into the state with
internal quantum numbers I] (decay channel 1]). Obviously one has
(6.4)
This statement is identical to the time derivative of (2.10).

Most often V is rotationally symmetric, so that only the term with j = lR

in (6.3) is different from zero.
The weight function p~(E) is related to the normalization of the generalized
eigenvectors IE b,jj31]):
(Ejhl]lE'j'j~I]') = b~~.(p~(E»-lb(E - E')bjj,b hA (6.5)
Instead of the basis vector (6.1) one may use different basis vectors, e.g.,
the generalized momentum eigenvectors
(p I] Ip' 1]') = b(p - p')b~~, (6.6)
if
[Pi' I]0P] = O. (6.6a)
Then one has instead of (6.3)
r(R ~ 1]) = 2n fd 3p b(E(p) - E R) I 2[1-1 I(p I] IV IER IR 13 KR) 12. (6.7)

13 R +
Instead of the momenta or angular momenta one can choose any other labels
IEb,b) = IE b, {3, 1]). (6.8)
The only requirement is that
For the basis vectors that one uses in the decay-rate formula one should
always choose the most convenient ones, and which the most convenient
ones are depends upon the particular problem. 24 Therefore we now discuss
the general case.
Let
(6.9)
be a complete system of commuting operators for the decaying system,
where we have split aOP = (KOP, o:oP) so that the decaying system has a definite
value K = KR for the first set of operators and o:OP have not been measured.
The reduced matrix element is then
(all WRIIQ') = (K 0:11 WRIIK' (1') = bKKR bK'KRbaa' dim ~(KR) (6.l0)
where ~(KR) is the eigenspace of K OP with eigenvalue KR.

Let
(6.l1 )
be a complete system of commuting observables chosen so that the detector
detects decay products with definite quantum number I] and detects states
with any values of {3. Then one can ask for the partial decay rate for the decay
24 For instance, if it should happen that (6.6a) is not fulfilled but that
Pi = PiM- 1
commutes with I)0P, where M is the mass operator (which is a function of the internal quantum
numbers), then one would choose the basis vectors Ip, I), where Pi = pJm(I)).
Problems 569
of R into the decay channell]. Inserting (6.10) into (5.14) and using eigen-
vectors of (6.9) and (6.11), one obtains
r = L r(R ~ 1]), (6.12)
where the partial decay rates are:
r(R f
~ 1]) = 2n p~(E) dE b(E - ER) t I<E I] [31 VIER KR aW. (6.13)
Here
L=
,p
1 LL
dim ~(KR) p ,
(6.14)
means summing (or integrating) over all values of the labels for the basis
vectors of the space of final states and averaging over all initial quantum
numbers. One often states this as "summing over the final states and
averaging over the initial states."
The formula (6.13) for the decay rates has numerous applications in all
branches of quantum physics. Whenever it makes sense to speak of a de-
caying quantum-mechanical system, (6.13) with suitably chosen basis vectors
can be applied. The interaction Hamiltonian or transition operator V
depends, of course, upon the particular physical system that one considers
and is determined by the algebra of observables. To find the right expression
for V is one of the tasks of understanding the physical system.
Problems
1. The decaying state vector IjD) is defined by
where IE +) are generalized eigenvectors of H:
HIE+) = EIE+).
Calculate the norm of the vector H IjD) for finite values of 2ER/f and for ER/f .... 00.
Interpret the result.
2. Show that the state WR given by (3.10) with (3.13) corresponds to a state of
exponential growth (" capture " state).
3. In this problem we suggest the calculations of the results in Section XXI.5 using
a different method.
(a) Define the quantity
y; Eb(E, E') = I I p(E)p(E')p(Eb)(Eb hi VIE a+)
bali'
x (E' a'+ WIEb h) (all WRlla')(p(ER»-l. (P.O)

Extend this function to complex values Wb, W, w' of the energy and show that,
as a consequence of the Hermiticity of Vand time-reversal in variance,
ff w.(W, w') = lJ Wb( ru, ru'), (P.la)
ff w.(w', w) = Y;Wb(W, w'). (P.1b)

Justify
using the symmetry relation of the S-matrix (Equation XVIII.5.lO). [(P.2)

is needed for the calculations of the following problems.]
(b) Use (3.10) with (3.14) to show that the decay rate (XIV.2JS) can be written as
(P.3)
with
e- iE1 eiE'1 1
x (PA)
E - ER + i(fj2) E' - ER - i(fj2) E' - Eb - iE'
8IW) = i ~ fIf dEb dE dE' g;d E, E')

e- iE1 eiE'1 I
x E _ ER + i(f/2) E' - ER - i(fj2) E - Eb + iE' (P.5)
and show that
(P,6)
(c) Use the approximation (3,7) and show, applying the Titchmarsh theorem, that
d()-4
q>t - n
2 r-r(2:Y
e Re,#'E R +iW/2l-i,
(.f
ER ,r)
-I2',E +I2" R (P.7)
(d) Show that from the condition q>(t -+ 00) = 1 it follows that
4nz2 Re ffER+iW/Zl-i,( ER - i~, ER + i~) = 1, (P.S)

and obtain (3040) and (3.41).
4. Assume that the detector has a very good energy resolution t1Eb <if r, Then the
decay rate per unit energy as a function of Eb can be measured.
(a) Using the same method as in Problem 3, show that this decay rate per unit
energy is given by
d# r
-E
d b (t = 0) = 2nf 2 Z ffEb(ER - iO, Eb + iO),
(Eb - ER ) + (r/2)
(b) Show that for the decaying angular-momentum state (3.16) the decay proba-
bility per unit energy is given by
d#(t = 0)
-...,-- = b(E h )
"2npb(Eb)"
1...
1 ~
--I<Ebbl VIER lR 13 KR)I ,
1... --
2
dEb h 1321R+1
where
b(E) = ~ 1
b 2n (Eb _-E-R--c)Z;-+-(f-/-2)-:Cz'
b(E b) is called the natural line width.

Hint: Use (P.2) of Problem 3.
Epilogue
The purpose of a physical theory is fulfilled if it provides a mathematical

image of some domain of reality that allows us to relate experimental data
and foresee new situations by making mathematical deductions. In this book
we have restricted ourselves to this program and have not mentioned any
of the further-reaching implications of quantum mechanics. Yet quantum
mechanics has affected all scientific and even general human thinking. And
when presented in its full generality, as done in this book, it effortlessly
reveals the two facts whose lessons reach far beyond the boundaries of
physics.
Classical science is based on two assumptions: (1) the deterministic
nature of predictions, and (2) the atomistic nature of understanding. Quan-
tum mechanics teaches us to revise both.
Classical theories are deterministic. The laws of classical physics are
constructed in such a way that if the initial values of a system's dynamical
variables are given, then their precise values can be calculated for any later
time. These laws were not simply derived solely from experience, and then
accepted as the basis of scientific philosophy and general thinking. Rather,
and perhaps to a greater degree, these laws have been extricated from nature
because tliey were in accord with the prevailing philosophical idea that
nothing can be without a cause.
Probability statements in classical physics are always associated with
insufficient knowledge, i.e., they are statements about the observer's know-
ledge and not about the physical system, which according to the principles
of classical physics can be known to unlimited accuracy. In quantum theory
571
572 Epilogue
(as described in Section U.5 and elsewhere in the text) one can only say with
what probability certain values can be expected, even if one knows the
state as well as possible, i.e., even if the system is in a pure state. Thus in
quantum theory statements are inherently probabilistic; the occurrence of
probability functions is not just a consequence of the observer's insufficient
knowledge, but a property attributed to the physical systems themselves.
Quantum predictions of experimental results are statements of how a micro-
physical process shows up in the macrophysical domain. These traces of
microphysical processes in the macrophysical domain, the only source of
human knowledge about such processes, do not obey deterministic laws.
Earlier traces of a microphysical process do not determine later traces
uniquely, but only probabilistically. Quantum theory teaches us that there
are inherent limitations to human knowledge.
The second point, that of the profoundly holistic nature (of the under-
standing) of quantum physical systems, is not often emphasized, even though
it is an obvious consequence of the quantum-mechanical description of
physical systems. Although holism has already become rather widely ac-
cepted in other disciplines (e.g., psychology), it has been resisted by the
physicists, who seem to be influenced by the success of atomism in classical
physics. The quantum physical system is a structured whole described by
the mathematical structure of an algebra of operators. From the laws of
the combination of quantum physical systems (Section III.5), it follows that
there are observables-ofthe combination of the two subsystems (described
by :Yf 1 Q9 :Yf 2) that are incompatible with all observables of either subsystem
(described by :Yf 1 or Yf 2)'
Thus, in quantum physics there exist holistic properties that cannot be
obtained as combinations of the properties of the subsystems. In this sense
the whole is not the sum of the parts.
In the atomistic approach understanding comes from the reduction of
the complex system to simpler subsystems by ever finer separations until
one comes to the ultimate constituents. In quantum physics the presence of
holistic properties prevents this reduction process, and the notion of ul-
timate constituents loses its meaning. Atomism belongs to classical physics.
A quantum physical system such as a molecule cannot be fully understood
by dissecting it into nuclei and electrons, although, in the tradition of our
scientific heritage, it is tempting to do this. What one arrives at in this way,
however, is only the classical analogue of the quantum physical system, as
in the Kepler system of proton and electron for the classical analogue of
the hydrogen atom. An electron in an atom" is" something different from
an electron in a linear accelerator, and the whole picture of the electron can
only be displayed by giving its different aspects as they are mathematically
described by the various basis systems in the space of physical states.
The visual picture that one usually requires for the process of under-
standing is in quantum physics not the geometrical picture of the object,
but the picture of its image in the space of physical states. The reduction
from the more complex to the simpler is performed not on the physical
Epilogue 573
object, leading to simpler constituent objects, but on the space of physical

states, leading to the irreducible subspaces for ever simpler structures. At
every stage of this reduction one still has a whole picture describing all
aspects by the various basis systems of the subspaces. But within a subspace
the structure is simpler and describes a narrower domain of physics.
Dissecting a quantum physical system may destroy it. Therefore a quan-
tum physical system (such as the CO molecule) cannot be understood only
atomistically (as a di-atom) but is often more adequately understood
holistically and functionally (as a vibrator rotator).
Atomism has been a great achievement of the past, and most technology
is based on it. But quantum theory has revealed its limitations and shown
that even for simple systems a holistic method is needed also.
Bibliography
The cited literature has been chosen rather arbitrarily. I have not made a
systematic search for the most suitable list of books for further or sup-
plementary reading, and mention just those I happen to have come across.
Many of these books I have used myself.
The books and a few review articles are separated into several categories
and then listed in alphabetical order; a few comments are added here and
there.
1 Foundations of Quantum Mechanics

P. A. M. Dirac, The Principles of Quantum Mechanics, Clarendon Press, Oxford, 1958
(fourth edition).
Written by one of the creators of quantum mechanics many years ago, this is still
modern and one of the greatest books written on this subject.
J. M. Jauch, Foundations of Quantum Mechanics, Addison-Wesley, Reading, Mass.,
1968.
This text is concerned with the conceptual foundations of quantum mechanics
and contains practically no applications. Furthermore, it differs from the present
presentation by starting from the lattice structure of quantum mechanics. Still,
it is recommended even to those who do not want to learn lattice theory.
G. Ludwig, Grundlagen der Quantenmechanik, Springer, Berlin, 1954.
J. von Neumann, Mathematical Foundations of Quantum Mechanics, Springer, Berlin,
1932; Princeton University Press, Princeton, 1955.
This is the first book written on the Hilbert-space formulation of quantum
mechanics developed by its author.
574
Bibliography 575
2 Scattering Theory
A. I. Baz, Ya. B. Zeldovich, A. M. Perelomov, Scattering Reactions and Decay in Non-
relativistic Quantum Mechanics, Israel Program for Scientific Translations,
Jerusalem, 1969.
L. Fonda, G. C. Ghirardi, A. Rimini, Decay Theory of Unstable Quantum Systems,
Reports on Progress in Physics, 41,587 (1978).
M. L. Goldberger, K. M. Watson, Collision Theory, Wiley, New York, 1964.
R. G. Newton, Scattering Theory of Waves and Particles. McGraw-Hill, New York,
1966.
In particular, for the formal theory of scattering this book is highly recommended.
H. M. Nussenzveig, Causality and Dispersion Relations, Academic Press, New York,
1972.
A. G. Sitenko, Lectures in Scattering Theory, Pergamon Press, New York, 1971.
K. Smith, The Calculation of Atomic Collision Processes, Wiley-Interscience, New York,
1971.
John R. Taylor, Scattering Theory, Wiley, New York, 1972.
This is a very clearly written book, which contains more material than the second
part of the present book, however, it is written in a somewhat different spirit. It
is highly recommended.
3 Theory of Angular Momentum and Group Theory

L. C. Biedenharn, J. D. Louck, Angular Momentum in Quantum Physics, Addison-
Wesley, Reading, Mass., 1979.
A. R. Edmonds, Angular Momentum in Quantum Mechanics, Princeton University
Press, Princeton, 1957.
I. M. Gelfand, R. A. Minlos, Z. Ja. Shapiro, Representations of the Rotation Group and
of the Lorentz Group, Pergamon Press, New York, 1963.
M. A. Naimark, Linear Representations of the Lorentz Group, Pergamon Press, New
York, 1964.
The first part of this book gives an introduction to the theory of group representa-
tions and discusses the rotation group in full detail. This is a book written by a
mathematician for physicists, very readable, and mathematically rigorous. Some
of the material of Chapter I is contained in this book.
M. Hamermesh, Group Theory, Addison-Wesley, Reading, Mass., 1962.
L. Michel, Applications of Group Theory to Quantum Physics: Algebraic Aspects; and
L. O'Raifeartaigh, Unitary Representations of Lie Groups in Quantum Mechanics.
Both published in Group Representations in Mathematics and Physics (Ed. V.
Bargmann). Springer Lecture Notes in Physics, Vol. 6,1970.
The two reviews by Michel and O'Raifeartaigh discuss many applications of group
theory to quantum physics.
M. E. Rose, Elementary Theory of Angular Momentum, Wiley, New York, 1957.
576 Bibliography
4 Experimental Subjects
N. L. Alpert, W. E. Keiser, H. A. Szymanski, Theory and Practice of Infrared Spectro-
scopy, Wiley, New York, 1970.
A. Barbaro-Galtieri, Baryon Resonances, in Advances in Particle Physics, Vol. 2 (Eds.
R. L. Cook, R. E. Marshak) Interscience, New York, 1968.
R. P. Bauman, Absorption Spectroscopy, Wiley, New York, 1962.
G. Herzberg, Molecular Spectra and Molecular Structure, D. van Nostrand, New York,
1966 (in particular Vol. 1).
This book teaches more than just molecular physics. It is one of the most beautiful
books on physics and is highly recommended to every student.
H. S. W. Massey, E. H. S. Burhop, H. B. Gilbody, Electronic and Ionic Impact Phenom-
ena, Clarendon Press, Oxford, 1969 (in particular Vol. 1: Collisions of Electrons
with Atoms).
5 Mathematical Material
Milton Abramowitz and I. A. Stegun, Handbook of Mathematical Functions, Dover,
New York, 1965.
A. Bohm, The Rigged Hilbert Space and Quantum Mechanics, Springer Lecture Notes
in Physics, Vol. 7g, 1978.
A. Bohm, M. Gadella, Dirac Kets, Gamow Vectors and the Mathematics of Scattering
and Decaying States, Springer Lecture Notes in Physics, Vol. (1986), to appear.
I. M. Gel'fand and G. P. Shilov, Generalized Functions, Vols. 1,2,4, Academic Press,
New York, 1964.
A. Lichnerowicz, Linear Algebra and Analysis, Holden-Day, San Francisco, 1967.
K. Maurin, Methods of Hilbert Spaces, Polish Scientific Publishers, Warsaw, 1967.
K. Maurin, General Eigenfunction Expansions and Unitary Representations of Topo-
logical Groups, Polish Scientific Publishers, Warsaw, 1968.
V. I. Smirnov, A Course of Higher Mathematics, Pergamon Press, New York, 1964 (in
particular Vol. III, Part 1, on Linear Algebra, and Vol. III, Part 2, on Functions
of a Complex Variable).
G. N. Watson, Theory of Bessel Functions, Cambridge University Press, Cambridge,
1958.
6 History of Quantum Mechanics

F. Hund, Geschichte der Quantentheorie, BI Wissenschaftsverlag, Berlin, 1975.
Max Jammer, The Conceptual Development of Quantum Mechanics, McGraw-Hill,
New York, 1966.
7 Quantum Mechanics, Textbooks and Atomic Physics

G. Baym, Lectures on Quantum Mechanics, Benjamin, New York, 1969.
H. A. Bethe, E. E. Salpeter, Quantum Mechanics of One- and Two-Electron Atoms,
Springer-Verlag, New York, 1957.
D. I. Blokhintsev, Quantum Mechanics, Reidel, Dordrecht, 1964.
Bibliography 577
R. H. Dicke, J. W. Wittke, Introduction to Quantum Mechanics, Addison-Wesley,

Reading, Mass., 1960.
R. P. Feynman, R. B. Leighton, M. Sands, The Feynman Lectures on Physics, Vol. 3,
Quantum Mechanics, Addison-Wesley, Reading, Mass., 1965.
S. Gasiorowicz, Quantum Physics, Wiley, New York, 1974.
D. T. Gillespie, A Quantum Mechanics Primer, Intex Publisher, 1970.
K. Gottfried, Quantum Mechanics, Benjamin, New York, 1966.
L. D. Landau, E. M. Lifshitz, Quantum Mechanics, Pergamon Press, New York, 1958.
H. J. Lipkin, Quantum Mechanics, New Approaches to Selected Topics, North-Holland,
Amsterdam, 1973.
E. Merzbacher, Quantum Mechanics, Wiley, New York, 1961.
A. Messiah, Quantum Mechanics, Vols. 1,2, Interscience, New York, 1961.
M. Mizushima, Quantum Mechanics of Atomic Spectra and Atomic Structure, Benjamin,
New York, 1970.
M. Sargent, M. O. Scully, W. E. Lamb, Laser Physics, Addison-Wesley, Reading, Mass.,
1974.
L. T. Schiff, Quantum Mechanics, McGraw-Hill, New York, 1968 (third edition).
8 Prerequisites from Classical Physics

V. D. Barger, M. Olsson, Classical Mechanics: A Modern Perspective, McGraw-Hill,
New York, 1973.
A. O. Barut, Electrodynamics and Classical Theory of Fields and Particles, Macmillan,
New York, 1964.
H. C. Corben, Classical and Quantum Theories of Spinning Particles, Holden-Day,
San Francisco, 1968.
H. Goldstein, Classical Mechanics, Addison-Wesley, Reading, Mass., 1980 (second
edition).
J. D. Jackson, Classical Electrodynamics, Wiley, New York, 1975 (second edition).
F. Rohrlich, Classical Charged Particles, Addison-Wesley, Reading, Mass., 1965.
Index
Absolutely continuous spectrum 16 operator *- 8, 161

Absorption spectrum-generating 216
of radiation 119 of the rotator 140
spectrum of helium 306 sub- 7
Addition with unit element 7
of operators 6 *- 8
of vectors 3 Algebraic
theorem for spherical conjugate space 34
harmonics 238, 413 dual space 34
Additive semigroup 311 Alkali atom 223
Adjoint Hamiltonian 223
M0ller operators 403 Almost
operator 6, 37 eigenstate 99
Advanced Green's function 398 sharp state 99
Algebra Amplitude
associative 7 background 490
enveloping 132 Breit-Wigner 483
generated 8 partial-wave 415
of 6'(E3) 182 elastic scattering 415
of 6'(SO(3)) 132 reaction 415
of 6'(SO(3, I)) 182 resonant 481
of 6'(SO( 4)) 182 scattering 413
of 6'(SU(2)) 132 elastic 384
of operators forward 386
continuous 36 Analytic
of the harmonic oscillator 45 continuation 446
of the rotator 138 function 444
579
580 Index
Analyticity properties of the Axial-vector 193

S-matrix 467,470 Axiom, basic (see Postulate)
Analyzer 55
Angular Background
momentum amplitude 490
basis 410, 567 integral 555
intrinsic 254 phase shift 480
operator 130 Balmer 220
commutation relations 131 Balmer series 220
total 166 Bargmann, Vo 205
orbital 255 Basis
quantum number 137 adapted to a Stern-Gerlach
total 255 apparatus 343
velocity 128 angular momentum 410,567
Anharmonic oscillator components of a vector with respect
energy levels 123, 124 to 11
infrared traOnsitions 123, 124 continuous 22
Annihilation operator 53 direct product 146, 148, 167
Anomalous generalized 87
magnetic moment 265 momentum eigenvector 410, 568
Zeeman effect 270 in an n-dimensional space 5
Antiadjoint operator 508 matrix of an operator with respect
Anticommutator 208 to 20
Antilinear physical 166
functional 34 statement for a 60
operator 6, 508 Bell 354
Antisymmetric Bell's inequalities 349-352, 354
subspace 284 Bessel function, spherical 430
vectors 280 Bohr
Antisymmetrical representation 279 magneton 266
Antiunitary 508 radius 231
Anzatz (hypothesis) 29, 48, 440 Bohr, Niels 47, 81,270,353
Argand diagrams 422,489-491, 532, Born approximation 380, 381, 385,
535 386, 405, 424, 566
Associated Born, ~ax 47,48,80,81
Laguerre polynomials 239 Bose statistics 280
Legendre polynomials 238 Bosons 280
Associative algebra 7 Bound
Assumption state poles 440,475, 550
basic (see Postulate) states 360, 390, 439
reproducibility 348 space of 211
Asymptotic subspace of 390
behavior of the lth partial S-matrix Bra 9
element 441 Branch
completeness 404 cut 449
Asymptotically complete 390 line 449
Atomism 572 points 445, 449
Attractive Branches 449
in an interval 433 Breit - Wigner
force 433 amplitude 483, 542
potential 433 approximation, generalized 522
Average cross section 503
time delay 462 curves
value 62, 76, 345 interfering 531
Averaging over polarization states noninterfering 531
381 formula 483
Index 581
Canonical commutation relations, Commutator 7

three-dimensional 128 Commuting operators 7, 69
Capture state 545, 569 Compatible observables 72
Cascade transition 351 Complete system
Casimir operators of commuting observables
of t&'(SO(4» 212 (c.s.c.o.) 160,388
of t&'(SU(2» 133 of eigenvectors 11
Cauchy Completeness
inequality for sequences 78 asymptotic 404
integral formula 31, 444 property of a basis 360
residue theorem 446 Complex eigenvalue 12, 18,547,
-Riemann conditions 444 562
-Schwartz- Bunyakovski Compliment, orthogonal 390
inequality 5 Component of a vector with respect to
Causality 464, 466, 467 a basis 9-11
Centrifugal Composite system 352
distortion 144 Conjugate
forces 145 operator 559
Change of state 72 space 34
by the dynamical law 328 Conservation of energy 379,401
by the measuring process 328 Constants of the motion 211,315
Channel Continuous
excitation 418,419 basis 22
inelastic 418,419 functionals 36
ionization 418,419 operator 37
rearrangement 419 operator function 325
Classical harmonic oscillator 46 at a point 324
Clebsch-Gordan coefficients 167 spectrum 11, 12, 16
explicit expressions for 172 discrete eigen val ues in 12
formula for 172 Contraction, Inonii-Wigner 189
orthogonality relations 167 Convergence
phase convention 170 Hilbert space 35
CO molecule 55,57, 121 of infinite sequences 35
Coefficien t radius of 445
Clebsch-Gordan 167 stronger 36
Fourier 25 weaker 36
inelasticity 421 <l>-space 36
transition 21 Convolution 378, 397
Wigner 167 Coordinate of a vector with respect to a
Coherent basis system 9-11
combination 73 Copenhagen interpretation 81
state vector 112 Correlated photon pairs 351
Collision Countably normed space 324
cross section, total 364 Coupling
rearrangement 419 constant 152
Combination rule for spherical harmonics 238
coherent 73 Creation operator 53
incoherent 73 Cross section
linear 5 Breit-Wigner 503
of two elementary rotators 165 classical 363
of two quantum-mechanical collision, total 364
systems 147 differential 363-365, 367, 382, 383,
interaction free 148 491
Commutation relation elastic 523
canonical 47 total 365
Heisenberg 47 energy-averaged 498
582 Index
Cross section (continued) Density matrix 64, 95

formula 378 Derivative of an operator 325
for coarse-resolution Detailed balance, principle of 514
experiments 379 Deterministic 328, 571
inelastic 365 Diagonal form, transformation of a
Ith partial 414,421 matrix into 22
elastic 421 Diatomic molecule, vibrating, rotating,
inelastic 421 interacting 152
observed 500 Differentiable operator function at a
partial 364 point in a subspace 325
production 363 Differential cross section 365
pure resonance 482 Dimension of a subspace 66, 67
quantum mechanical 364, 368 Dipole
reaction 523 moment for the helium atom,
scattering 363 operator for 297
differential 363, 364 resonance 487
total 363 transitions 120,216
total Dirac
inelastic 365 delta-function 15
inelastic partial picture 321
for the 11th level 365 Dirac, P. A. M. 9, 80, 160, 566
e.s.c.o. 160, 388 Direct
Cyclic 11 product (see Tensor product)
basis 146, 148, 167
of operators 147, 157
Davisson 107, 116 state 331, 352, 369
De Broglie, Louis 48, 107 -product space 146
De Broglie relation 104, 209 sum 59
Decay of operators 157
probability 540 Discrete
for a quasistationary state 565 eigenvalues in the continuous
processes 538 spectrum 12,390
rate 539, 565 spectrum 11
initial 539, 549 Dispersion
partial 567 of an observable in a state 76
per unit energy 565 relations 467
total 566 Distribution 15, 27
Decaying state 18, 346, 537, 538 Double
Decomposition of a space into ionization threshold of the helium
orthogonal subspaces 59 atom 307
Defining algebraic relations 8 multichannel resonance 520, 526
of the generators of the algebra of cross section of 526
angular momentum 166 resonance 520
of the harmonic oscillator 45 scattering 303
of the spin operators 131 Doubling, parity 195
Degenerate Doubly excited states of helium 308
case perturbation theory 243, 252 broad profile of 309
eigenvalue 66,217 Dual space 34
spectrum 11, 66 Dumbbell 129, 131, 198
Degenerate case perturbation
theory 243
De Haas 265 tff(E 3 ) 182
Delta-function tff(SO(3» 132
normalization 87 tff(SO(3, 1» 182
Dense subspace 38, 326 tff(SO(4» 182,212,213
Index 583
,g'(SU(2» 132 Elastically scattered 57

(21 + I)-dimensional irreducible Elasticity 533, 536
representation space of 132, Electromagnetic spectrum 122
136 Electron 253
Effective Electronic states of a molecule 155
quantum number 231 Electrostatic units 206
range Elementary
approximation 443 particle, free-nonrelativistic 103
of a potential 443 rotator 164, 165
Efficiency function, detector 498 Emission of radiation 119
Eigenfunction Empirically 350
expansion 23, 24 Energy
generalized 27 conservation 379,401
of an operator 23 diagram, symmetric top 201
Eigenfunctions, orthogonal 24 distribution
Eigenkets II Gaussian 486
Eigenphase 518 Lorentzian 486
repulsion theorem, Wigner's 520 -level diagram 57
-shifts 518 of the harmonic oscillator 57
Eigenspace of an operator 63, 65, 66, rotation-vibration band 150
77 with P, Q, and R branches 202
invariant 211 level of the hydrogen atom, n 2 -fold
projection operator on 68 degeneracy of 217,220
Eigenstate of an observable 77 loss 57
almost 99 -loss
physical meaning 77 experiments 55
pure 62 spectrum of helium 305
Eigenstates of the energy operator, operator 313
experimental verification that experimental verification that the
the physical states are 303 physical states are eigenstates
Eigenvalue 6, 72 of 303
complex 18, 547 for the harmonic oscillator 45, 50
degenerate 66 spectral representation 61
generalized 38 for the hydrogen atom 208
problem in matrix notation 22 kinetic (see Kinetic energy
proper 38 operator)
Eigenvector of the quantum-mechanical
expansion 10, 11 symmetric top 199
generalized 11, 18 resonance 482, 535
nondegenerate 11 shell 379
generalized 11,38, 162 spectrum
of a linear operator 6 experimental 58
proper 38, 546 of the harmonic oscillator 54
Eigenvectors of the rotator 140
basis system of 11 of the symmetric top 201
complete system of 11 Ensemble 55
Einstein, Albert 80, 107, 265, 270, 353, state of an 58
354 Entire function 446
Elastic Enveloping algebra 132
channel 418,419 Environment, measurement processes
cross section 523 caused by 539
total 365 EPR paradox 353
scattering 384 Equation of motion, Heisenberg 315
amplitude 384, 422 Equilibrium, thermal 381
partial-wave amplitude for 415 Essential singularity 445
584 Index
Essentially self-adjoint 6, 38, 69 Force

Euclidean attractive 433
group, three-dimensional 182 centrifugal 145·
space 4 Lorentz 263
Even permutation 276 operator 316
Exact repulsive 433
eigenvector 242 Formation
equation 290 experiments 454
Green's functions 398 process 537
Hamiltonian 242 Forming state 538
radial wave function 427 Forward scattering amplitude 386
state vector 392 Fourier
wave function 427 coefficients 25
Excitation energies of helium 303 expansion 25
Excited, doubly 308 series representation 25
Exclusion principle 280 transform 26, 353, 396
Expansion inverse 26, 396
eigenvector 10 of the unit step function 397,461
Fourier 25 Fraction of a resonance 533
Expectation value 64, 70, 79, 345, 346 Franck 55, 80
in a pure state 62 Franck-Hertz experiment 55
Experimental Frechet-Riesz theorem 37,87
data 49 Free
energy spectrum 58 eigenvector 242
verification that the physical states Green's functions 398
are eigenstates of the energy Hamiltonian 242
operator 303 -nonrelativistic elementary
Experiments particle 103
actual 349 radial wave function 427
conceivable 349 state 391
formation 454 vector 391
gedanken (see Gedanken experiment) wave function 427
production 454 Frequency 121
Exponential decay law 540 Function
deviations from 549 analytic 444
Extension branch of 449
of an operator 25, 38 Dirac delta- 15
of the scalar product 37 eigenfunction expansion of 23
of the unitary time development entire 446
operator 562 generalized 15
generating, of the Hermite
polynomials 32
meromorphic 446
Factorial, double 430 in a region 445
Fano shape parameters 494 multivalued 448
Fermi statistics 280 normalized 15
Fermions 280 of a complex variable 444
Final state 366 of operators 69, 88
Fine-structure 124, 151, 162, 221 sequences of delta-type 97
constant, Sommerfeld 209, 265 square integrable 17
interaction 269 well behaved 17
splitting 269 Functional 33
Finite anti linear 34, 86-88
-dimensional space 5 continuous 36
spectral theorem for 10 linear 34
group 276 value of 37
Index 585
Galilei transformation 111 symmetrical representation of 279

Gamow unitary representation of 276
kets 542, 557 SU(2), generators of 132
vectors 542, 557 symmetric 276
Gaussian energy distribution 486 Gyromagnetic ratio 262
Gedanken experiment (too idealized
to be performed in practice,
and the results of which can Hif -class function 554
only be obtained by thought Hamiltonian
based on many other exact 242
experiences) 78, 79, 81, 328, free 242
329,336,463 Green's function for the operator
Gelfand, I. M. 81 for 243
Gelfand operator 313
- N aimark -Segal reconstruction resolvent of the 243
theorem 8 perturbation 242
triplet 37 Hankel functions, spherical 431,471
Generalized Hardy class functions 553
basis vector expansion 556, 562 from above, space of 553
Breit - Wigner approximation 522 from below, space of 553
eigenvalue 12, 38 Hardy class vector from above 554
eigenvectors 11, 12, 38 Harmonic oscillator
eigenvalues corresponding to 38 algebra of operators 45
complex 18 classical 46
normalization 16 energy spectrum of 54
function 15 one-dimensional 44
optical theorem 406,418 classical mechanical 45
Generated algebra 8 selection rule for 120
Generator 312 wave-functions 89
of time evolution (see Heisenberg
Hamiltonian) 313 commutation relation 47
Generators equation of motion 315
of the group SU(2) 132 picture 315
set of 8 uncertainty relation 77
Gerlach 336 Heisenberg, Werner 47,48,81,270,
Germer 107, 116 287, 366
Gibbs distribution 71, 381 Helicity 140
Golden Rule 563, 566 Helium atom 282
Goudsmit, S. 270 absorption spectrum of 306
Green's function doubly excited 308
advanced 398 energy operator of 283
exact 398 excitation energies 303
free 398 space of physical states 285
for the Hamiltonian, operator for Hermite
the 243 differential equations 30
Ith-partial-wave 434 functions, recurrence relations
retarded 398 for 86
Ground state 58 polynomials 30, 86
Group generating function of 32
finite 276 orthogonality relation for 30, 89
of operators, one-parameter 312 recurrence relation for 33
orthogonal, in four dimensions Rodrigues formula for 30
182 Hermitian
permutation 276 form 4
antisymmetrical representation posItIve 4, 84
of 279 positive definite 4
586 Index
Hermitian (continued) Inonii-Wigner contraction 189

operator 6, 7 Integral transform 28
vector 181 Interaction 148, 152
Hertz 55 free state 369
Hidden variables 348, 353 picture 322
local theories 354 Interference 74
theories 354 Interfering Breit - Wigner
Hilbert, David 80 resonances 531
Hilbert space 17, 54, 324 Interpretation, Copenhagen
convergence 35, 325 (statistical) 81
pre- 4 Intertwining relations 401
realization of 17 Intrinsic
rigged 37, 87, 99, 559 angular momentum 49, 254
Holism 572 parity 195, 197
Huygens, C. 106 Invariance, parity 260
Hydrogen atom Invariant
energy levels, low-lying 221 operator (see Casimir operator)
energy operator for 208 space 211
Hyperfine Inverse
interaction 272 Fourier transform 26, 396
splitting 222 of an operator 7, 208
Involution 8
Ionization
Identical quantum-mechanical channel 418
particles 275 energy 296
Identity operator 6 -excitation channels 419
spectral resolution of 13, 69 potential 296
In-state 393, 395, 551 threshold
Incoherent combination of states 73 nth 292,307
Incompatible observables 72 of the helium atom 307
Independent, linearly 5 Irreducible representation space
Indistinguishability 277 of 6'(SO(4)) 212
Indistinguishable 275 of 6'(S U(2)) 136
particles 274 Isolated
statement that identical particles physical systems 311, 318, 323, 539
are 277 singularities 445
Indivisible whole 352 Isometric
Inelastic operator 402
channel 418 scalar product spaces 5
cross section 365 Isomorphic
process, threshold of 418 linear spaces 5
reaction processes 417 scalar product spaces 5
Inelastically 57 Isomorphism 5
Inelasticity 523
coefficient 421, 533
Infinite
-dimensional space 5, 10 Jordan, P. 48, 80
sequence, square summable 14
Infinitesimal operator 312
Infrared spectrum of HCI 123, 126 Kepler problem
Inhomogeneous magnetic field 340, classical 206
342 quantum-mechanical 205,208
Initial Kernel of an integral transform 28
decay rate 539 Ket 9
state 366 Gamow 542, 557
Index 587
Kinetic energy operator -wave Green's function 434

for a massive nonrelativistic Lyman series 220
particle 370
for a photon 370
Kronecker delta 13, 15 Magnetic field, inhomogeneous 340,
Kronig, R. 270 342
Magnetic moment
anomalous 265
Ladder of a neutron 265
operators 53 of a proton 265
representations 133, 188 of a spinning particle in classical
Laguerre polynomials 239 physics 262
Lamb shift 269 of an electron 264
Lande 270 operator of a particle without intrinsic
Lande factor 262 magnetic moment 264
Larmor Magneton, Bohr 266
frequency 341 Many-particle systems 274
precession 340, 341 Mass
Laurent series 445 operator 568
Legendre polynomials 238,501 reduced 46, 129
orthogonality property of 414 Mathematical image of a physical
associated 238 system 44, 100
Lenard, P. 107 Matrix
Lenz vector 207, 209 elements 21
quantum-mechanical 209 of a map 40
Level shift 244 of an operator 20
operator 243 orthogonal 22
Lifetime 540, 549 symmetric 514
Line width, natural 565, 570 unitary 22
Linear Mean life 540
combination 5 Measurement 66, 76, 339
functional 34 change of state by 72, 328, 338, 539
operator 6, 508 immediate repetition of 65, 66
eigenvalue of 6 numbers obtained 7, 39
eigenvector of 6 of spin components 342
space 3 process caused by the
Stark effect 252 environment 539
topological space 324 Meromorphic function 446
Linearly in a region 445
dependent 5 Metastable states 454
independent 5 Microreversability, principle of 514
Lippmann-Schwinger equation 250, Minimal coupling to the
359,387,426 electromagnetic field 265
Local hidden variables theories 354 Mixing, singlet-triplet 299
London,F. 80 Mixture 57, 58, 63, 72
Lorentz Meller wave operators 400
force 263 adjoint 403
group 182 Moment of inertia 128
Lorentzian Momentum
energy distribution 486, 543 operator
momentum distribution 115 for the harmonic oscillator 45
lth partial radial 228
cross section 414,416,421 representation, wave functions in 90
S-matrix element, symmetry relation space wave function 91
for 472,474 transfer 384
588 Index
Monochromator 55 Odd permutation 276

Motion, Heisenberg equation of 315 Off the energy shell 379
Multichannel resonance 517 Old quantum theory 47
double 520 On the energy shell 379
single 522 One-parameter group of
experimental examples 525 operators 312,325
Multiplication One-particle systems 274
of an operator by a number 6 Operator
of two operators 6 addition 6
Multipole transitions 216 adjoint of 6, 37
Multivalued function 448 angular momentum 130
Mutually orthogonal subspaces 59 total 166
antiadjoint of an 508
antilinear 6, 508
antiunitary 508
n-dimensional space 5
axial-vector 193
basis in 5
Casimir 133, 212
Natural line width 565, 570
commuting 7, 69
Negation of a property 109
Negative-energy space 211 conjugate 559, 562
Neumann function, spherical 430 continuous 37
Newton 47, 106 derivative, at a point 325
Newton's equation, quantum- eigenfunctions of 23
mechanical 316 eigenspace of 66
Nondecay probability 540 eigenvalue of 72
Nondegenerate spectra 159 generalized 38
Nonrelativistic symmetric top, extension of 38
classical 199 for the Green's function for the
Nonresonant background, slowly Hamiltonian 243
varying 496 function 69, 324
continuous 324, 325
Nordheim, L. 80
differentiable 325
Norm 4
infinitesimal 312
Normalized 4
inverse of 7, 208
delta-function 15
isometric 402
Normed space
countably 324 ladder 53
nth ionization threshold 292 level shift 243
Nuclear spectral theorem 10, 16, 162, linear 6, 508
163 matrix of 20
Moller wave 400
adjoint of 403
multiplication 6
Observable 61 by a number 6
dispersion of 76 positive 58
expectation value of 64 positive definite 208
in a pure state 62 product 7
experimentally determined value projection 13, 58
of 61 proper tensor 193
mathematical image of 62, 71 pseudotensor 193
physical 44 regular tensor 176
pure eigenstate of 62 scalar 176
values of 71 self-adjoint
Observables complex eigenvalue of 18, 547
complete system of commuting spectral resolution of a 13, 16
(c.s.c.o.) 160 semibounded 471
incompatible 73 semilinear 44, 508
Index 589
semiunitary 509 intrinsic 195, 197

skew-Hermitian 313 invariance 260, 505
spectral family of 69 -invariant 505
spin 138 nonconservation 193
square root of 208 operation 192
statistical 64, 65, 83, 368, 548 operator 192
tensor for a two-electron system 298
ofrankj 176 relative 217
regular 176 Partial
trace of 59 cross section 364
unit 6 lth 414
unitary 7, 192,402 decay rate 567
vector 138, 176 -wave
well defined 40 amplitude 415
spectral representation of 69 for elastic scattering 415
*-algebra 8, 161 resonant 481
Operators reaction amplitude 415
algebra of, of the harmonic width 522
oscillator 45 Particle 102, 103
direct product of 157 elementary 106
direct sum of 157 free-nonrelativistic 103
one-parameter group of 312, 325 extended 131
Optical theorem 422 Particles
generalized 406, 418 identical 275
Orthogonal indistinguishable 274
basis functions 25 Pauli
decomposition of a space into exclusion principle 280
subspaces 59 matrices 138
eigenfunctions 24 Pauli, W. 205, 270
matrix 22 Permutation 275
subspaces, mutually 59 even 276
vectors 4 group 276
Orthogonality property of the Legendre antisymmetrical representation
polynomials 414 of 279
Orthohelium 287, 296, 297 symmetrical representation of 279
Orthonormal unitary representation of 276
basis odd 276
functions 29 operation 276
system 9 Perrin, 1. 107
on an interval 32 Perturbation 242
Orthonormality of spherical Hamiltonian 242
harmonics 460 of the continuous spectrum 248
Oscillator, harmonic 44 series
wave functions 89 Rayleigh-Schrodinger 247
Out-state 393, 395, 551 Wigner-Brillouin 247
theory 225, 242
degenerate case 243, 252
Phase
P branch 124, 149, 202 factor 30
p-representation, wave function in 91 shift 437
Paley-Wiener Theorem 554 analysis 489
Parahelium 287, 296, 297 background 480, 501
Parity 195 of scattering 420
conservation 300 potential part of 480
doubling 195 resonant part of 480
590 Index
Physical Pre-Hilbert space 4

basis 166 Precession, Larmor 340, 341
observable 37, 39,44 Preparation
sheet of the S-matrix 473 apparatus 74
states 3, 138, 552 of a state 65, 72, 137, 346, 352
preparable 74 Principal
space of 99, 210 quantum number 215
systems 8 -value integral 249
conservative 318 Principle
isolated 318 of detailed balance 514
mathematical image of 44, 100 of microreversability 514
theories of 124, 571 of stability 492
Picture Probabilistic 572
Dirac 321 Probability 65
Heisenberg 315 amplitude 105
interaction 322 density 80
Schrodinger 314,315 interpretation 80
Planck 107 observable 65, 70
Planck of decay 540
- Einstein relation 209 of nondecay 540
formula 107 of transition 358
Planck's constant 45, 48, 110, 112, 209 per unit time 368
Plane wave 103 Product of a vector with a number 3
Podolsky 353 Production
Polarization 368, 381 cross section 363
Polarized 328 experiments 454
Pole Profile index 494
bound state 440,475, 550 Projection operator 13, 58, 88
of order n 445 on an eigenspace 68
simple 445 Projector 58
Position operator for the harmonic Proper
oscillator 45 eigenvalue 38
Positive eigenvector 11, 38
definite tensor operator 193
Hermitian form 4 Property 65
operator 208 negation of 109
Hermitian form 4 Proposition 65
operator 58 Pseudo-orthogonal group in
Postulates, basic (3 + I)-dimensions 182
I 44 Pseudotensor operator 193
II' 62 Pure
II 64,346 eigenstate 62, 76
IlIa 67,99 state 58, 278
IIIb 72, 100, 339 mathematical image of 39, 61, 62,
IVa 147,343 64,67,72
IVb 280
Va 314
Vb 315
Vc 318 Q-branch 202
Potential Quantum
attractive 433 defect 231
in an interval 433 electrodynamics 221
effective range of 443 -mechanical
part of the phase shift 480 Lenz vector 209
repulsive 433 Newton's equation 316
Index 591
mechanical symmetric top 198 Representation 138

mechanics, some fundamental antisymmetrical 279
properties of 63, 78, 328 ladder 188
number 160 momentum, wave functions in 90
angular momentum 137 Schrodinger 93
principal 215 space
vibrational 141 irreducible 136
Quasistationary states 454, 464, 477, reducible 139
542 symmetrical 279
unitary 182,276
Repulsive
force 433
R-branch 124, 149,202 potential 433
Rabi 265 Residue of a function 446
Radial Resolution
momentum operator 228 of an apparatus 97, 99, 498, 499
wave function of the identity operator with respect
asymptotic behavior of 435 to a spectral family 16, 69
exact 427 Resolvant for the operator H 243
free 427 Resonance 309, 378,454,455,476,
Rate 485, 542
of decay 539 cross section, pure 482
of initial decay 539 double 520
of partial decay 567 energy 482, 535
of transition 358 fraction of 533
Ray 61 multichannel 517, 520, 522, 525
Rayleigh-Schrodinger perturbation poles 476, 478, 483, 485
series 247 total width of 536
Reaction tripole 504
amplitude, partial-wave 415 width 482
cross section 523 Resonant
processes, inelastic 417 part of the phase shift 480
Realization 3, 93, 165 partial-wave amplitude 481
of linear spaces 18 Rest energy 267
of operators 18 Retarded Green's function 398
Rearrangement Riccati
channel 419 - Bessel function 430
collision 419 -Neumann function 430
-excitation channels 419 Riemann
Reciprocity theorem 514 sheet 450
experimental check of 514 surface 449, 450
Recurrence relation Rigged Hilbert space 37, 87, 99, 559
for the Hermite functions 86 description 100
for the Hermite polynomials 33 Rigid rotator 128, 140
Reduced Rodrigues formula for the Hermite
mass 46, 129 polynomials 30
matrix element 177 Root-mean-square deviation 78
momentum 206 Rosen 353
Reducible representation space 139 Rotating-oscillator 146
Reduction of a space into a sum of Rotation 343
irreducible total angular angle 343
momentum spaces 170 axis 343
Regular tensor operator 176 group, quantum-mechanical 138
Relative probabilities 58 spectra 138
Relativistic mass effect 267 -vibration band 150
592 Index
Rotational Schrodinger's wave mechanics 80

absorption spectra of HCl 143 Schulz resonance 454, 456
constant of CO 152 Schwartz, L. 81
constant of H 2 molecule 152 Schwartz space 99, 439
constants of HCl 151 Schwarz reflection principle 447
invariance 257 Schwinger 265.
Rotator 139 Selection
algebra of operators 138 of a subensemble 72, 100
elementary 164, 165 rules 120, 270
energy spectrum 140 for dipole radiation of the
nonrigid 144 rotator 141
space of physical states of 138 for the harmonic oscillator 120
spectrum generating algebra of 140 for the helium atom 298
Runge-Lenz vector 207 for the hydrogen atom with
Rydberg constant 218,219 spin 270
Self-adjoint operator 6
complex eigenvalues of 547
S-matrix 366,399,404,416 essentially 38
pair of second sheet poles of 542 Semibounded operator 471
theory 366, 406 Semigroup, additive 311
S-operator 366, 399, 417 Semilinear operator 44, 508
Scalar Semi unitary operator 509
operator 176 Set of generators 8
product 4 Shape
continuous case 14 parameter, Fano 494
extension of 37 profile parameter 494
product space 4 Sheet
Scattered elastically 57 physical 473
Scattering Riemann 450
amplitude 413 unphysical 473
elastic 384, 422 Siegert poles 476,481, 542
forward 386 Simple
angle 384 pole 445
channels 419 spectra 159
cross section 363 Simultaneous
differential 363, 364 eigenstates 82
total 363 eigenvectors 133, 344
elastic 384 measurements 82, 344, 345
partial-wave amplitude for 415 Singlet
experiment 363 states 287, 296, 346
lengths 443 -triplet mixing 299
operator 399, 401 Singularities 445
phase shift 420 essential 445
potentials 432 isolated 445
states 218,360,390 Skew-Hermitian
space of 389, 390 operator 313
theory, formal 387 vector operator 181
Schrodinger Slowly varying nonresonant
equation 318, 319 background 496
for the one-electron atom 235 Solid angle delta-function 385
for the radial wave function 428 Sommerfeld fine-structure
time-independent 93 constant 209, 265
picture 214,215 Space
representation 93 antisymmetric 284
Schrodinger, Erwin 47, 48 conjugate 34
Index 593
countably normed 324 generating-algebra 216

decomposition of 59 of the rotator 140
direct-product 146 generating-group 216
dual 34 nondegenerate 159
Euclidean 4 simple 159
three-dimensional 59 Spherical
finite dimensional 5 Bessel function 430
Hilbert 17, 54 components of a vector
inversIOn 192 operator 176
irreducible representation 136 coordinates 236,371
linear 3 delta-function normalization
linear topological 37, 324 in 432
n-dimensional 5 Hankel functions 431,471
basis in 5 harmonics 237
negative-energy 211 addition theorem for 238,413
of bound states 211 completeness relation of 238
of continuous antilinear coupling rule for 238
functionals 11 orthonormality of 460
of continuous functionals 36 Neumann function 430
of Hardy class functions from symmetry 416
above 553 Spherically symmetric interaction 384
of Hardy class functions from Spin 255
below 553 and statistics, connection
of physical states 9, 99, 147, 210, between 280
325, 554 correlation observable 344
of the hydrogen atom 215 correlations in a singlet state 342,
of the quantum-mechanical 343
rotator 138 -down 332
of scattering states 390 electron 253
of square integrable functions 17, operators 131, 138
557 -orbit interaction 265, 300, 302
pre-Hilbert 4, 17 -up 332
reducible representation 139 Splitting
rigged Hilbert 37, 554 fine-structure 269
Schwartz 17,40, 439 hyperfine 222
self-dual 35 Square
spanned by a vector 59 integrable function 17
symmetric 284 summable infinite sequence 14
Spectral Stable states 454
decomposition of a vector 10 Standard deviation 78
family of an operator 69 Stark effect, linear 252
representation of an operator with State
discrete spectrum 61,69 almost eigen- 99
resolution almost sharp 99
of a self-adjoint operator 13, 16 capture 545, 569
of the identity operator 13, 69 change of 72
series of hydrogen 220 controlled 560
theorem 10- 12, 68, 162 decaying 18,537,538
Spectrum 10, 11, 54 discrete physical 58
absolutely continuous 16 dispersion of an observable in 76
continuous 11, 87 exact 392
discrete eigenvalues in 12 forming 538
degenerate 66 free 391
discrete 11, 87 ground 58
electromagnetic 122 in- 393, 395
594 Index
State (continued) Stern-Gerlach experiment 254, 329,

interaction free 369 342
mathematical image of 62, 64, 67 Stronger convergence 36
metastable 454 Structureless particles 55
momentum space wave function SU(2)
for 91 enveloping algebra of 132
nonstationary 357 generators of the group 132
of a combination of physical Subalgebra 7
systems 331,369 Subensemble, selection of 72, 100
of a system after a measurement 72 Subspace 5
of an ensemble 58 antisymmetric 284
of the target 369 dense 326
out- 393, 395 invariant 139
parity of 195 of bound states 390
physically preparable 74, 99 of scattering states 389
preparation of a 65 symmetric 284
pure 58, 62, 64, 67, 278 Subspaces, orthogonal 59
quasistationary 454,464,477, 542 Sum, direct 59
sharp 97 Superposition 73
singlet 287, 296 principle 74
stable 454 Symmetric
stationary 320, 454 (permutation) group 276
time-reversed 509 interaction, spherically 384
transition probability into 358 operator 6
value of an observable in 61 subspace 284
vector 61 top
exact 392 classical nonrelativistic 199
free 391 energy diagram 202
vectors, coherent combination of 73 quantum mechanical 198
velocity reversed 509 energy operator of the 199
virtual 475, 550 energy spectrum of the 201
Statement vectors 280
for a basis 60 Symmetrical representation 279
that identical particles are Symmetry
indistinguishable 277 breaking 231
States relation for the lth partial S-matrix
incoherent combination of 73 element 472,474
mixture of 58 relations of the Clebsch-Gordan
space of bound 211 coefficien ts 172
space of physical 99, 210 spherical 416
space of scattering 390 transformation 318
subspace of bound 390 System
subspace of scattering 389 energy operator of 313
triplet of 287, 296 Hamiltonian operator of 313
Stationary isolated physical 318
state 320,454 many-particle 274
system 320 mathematical image of 62
Statistical operator 64, 83, 95, 368 one-particle 274
choice of 65 stationary 320
of a decaying state 548
of a pure state 95
Statistics
Bose 280 T-matrix 367,379,416
Fermi 280 off-the-energy-shell 379
Stern 336 on-the-energy-shell 379, 413
Index 595
Target 369 Transition

Taylor series 444 amplitude 523
Tensor cascade 351
operator 176 coefficients 21, 23, 27
of rankj 176 dipole 216
proper 193 matrix 21
pseudo- 193 muItipole 216
regular 176 probability 358
product 146 per unit time 368
Term rate 358
diagram for energy levels of two-step 351
helium 295 Transpositions 276
value 153 Triplet states 287, 296
Thermal equilibrium 381 Tripole resonance 504
Thomas Two-electron excitation states in
factor 266 neutral helium 306
precession 266 Two-step transition 351
Thomas, L. H. 270
Thomson, G. P. 107
Thomson, J. J. 106 Uhlenbeck, G. E. 270
Threshold Uncertainty
nth ionization 292 of classical physics 354
of an inelastic process 418 quantum-mechanical 354
Time 310 relation, Heisenberg 77, 82
delay 458, 462 Unit
average 462 operator 6
dependence, explicit 323 step function 393
evolution 310 Fourier transform of 397,461
generator of 313 ray 61
-reversal 507 Unitarity of the S-operator 417
in variance 509 Unitary
operator 507 circle 422, 490
-reversed state 509 deficiency 403
Titchmarch Theorem 554 matrix 22
Top (see Symmetric top) operator 7,192,402
Topological representation
space 35 of E3 192
linear 324 of SO(3, 1) 192
structures 19, 35 of SO(4) 192
Topology 35, 36, 99, 325 of the permutation group 276
coarser 36 Units 109-112,209
finer 36 Unphysical sheet of the S-matrix 473
Torque 262,263
Total
-angular momentum operators 166 Value of an observable 71, 76, 344
collision cross section 364 average 62, 76, 345
decay rate 566 expectation 64, 345
elastic cross section 365 in a pure state 62
inelastic cross section 365 experimentally
inelastic partial cross section for the determined 61, 72, 76, 86, 344
11th level 365 produced 347, 353
scattering cross section 363 uncovered 347, 353
width of a resonance 536 Vector 2
Trace of an operator 59 addition 3
Transformation matrix 21 antisymmetric 280
596 Index
Vector (continued) packet 103, 550

axial- 193 particle-duality 81
components of II plane 103
exact state 392 Wave functions in the momentum
free state 391 representation 90
Gamow 542, 557 Weaker convergence 36
Hardy class Weight 133
from above 554 diagram 136
operators 138, 176 function 16, 568
Hermitian 181 highest 134
skew-Hermitian 181 vectors 133
spherical components of 176 Well-behaved functions 15, 17, 27, 40
product with a number 3 Well-defined operator 40
space spanned by a 59 Wheeler, J. A. 351, 366
spectral decomposition of 10 Width
state 61 natural line 565, 570
symmetric 280 of a resonance 482
value of a functional at a 37 total 536
weight 133 Wigner
Velocity-reversed state 509 -Brillouin perturbation series 247
Vibrating rotator 146, 148 coefficients 167
Vibration-rotation band of carbon - Eckart theorem 177
monoxide 124, 125 theorem 132,509,512
Virtual 3-j symbols 173-175
state 475, 550 Wigner, E. P. 138
-state poles 475, 487 Wigner's
Voigt integral 502 causality inequality 466
Voigt profile 502 eigenphase repulsion theorem 520
Von Kampen causality condition 464
Von Neumann, John 80,324
Young 106
Yukawa interaction 385
Wave 103
function 80, 105
exact 427 Zeeman effect 273
exact radial 427 anomalous 270
free 427 Zero operator 6
free radial 427 Zero vector 3
in the momentum
representation 90
momentum space 91 *-algebra 8
number 121 *-operation 8
operator, Moller 400 =2 214

(Arno Bohm (Auth.) ) Quantum Mechanics Foundations (B-Ok - Xyz)

Caricato da

Informazioni sul documento

Titolo originale

Copyright

Formati disponibili

Condividi questo documento

Condividi o incorpora il documento

Opzioni di condivisione

Hai trovato utile questo documento?

Questo contenuto è inappropriato?

Copyright:

Formati disponibili

(Arno Bohm (Auth.) ) Quantum Mechanics Foundations (B-Ok - Xyz)

Caricato da

Copyright:

Formati disponibili

Texts and

R. Bass: Nuclear Reactions with Heavy Ions (1980).

Second Edition, Revised and Enlarged

6 Springer Science+Business Media, LLC

Elliott H. Lieb Tullio Regge Walter Thirring

Library of Congress Cataloging in Publieation Data

© 1979, 1986 by Springer Science+Business Media New York

Typeset by Composition House Ltd., Salisbury, England.

ISBN 978-3-540-13985-0 ISBN 978-3-662-01168-3 (eBook)

very helpful. Prerequisites are a knowledge of calculus, vector algebra, and

... in contrast to what one finds in the standard books, quantum

1 The subject also is usually adequately treated in undergraduate courses.

This is a physics book, and though mathematics has been used

II.8 Postulates II and III for Observables with Continuous Spectra 94

The mathematical language of quantum mechanics is introduced in this

1.1 The Mathematical Language of Quantum Mechanics

Therefore we shall start the quantum-mechanics course with some of the

1.2 Linear Spaces, Scalar Product

A linear space is a set with very little mathematical structure. We will

Positive Hermitian forms, which are not necessarily scalar products,

1.3 Linear Operators

eigenvectors with different eigenvalues. There are also linear operators on

An algebra.s;1 is called a *-algebra if we have on the algebra a *-operation

1.4 Basis Systems and Eigenvector Decomposition

Their scalar products are written in one of the following ways:

i,j = 1,2, .... (4.2)

If <I> is N-dimensional then there are N linearly independent vectors in this

i.e., choose-according to (3.6a)-eigenvectors of the tensor L

The symbol I >with a letter <jJ, a or i in it denoting a vector <jJ, an eigenvalue

We will call the self-adjoint operator with a discrete set of eigenvalues H

The spectral theorem (4.4b), (4.4d), (4.4c), and (4.4g)-which we have to

(IE m), cf» = L (IE m), IEn»(En Icf». (4.6d)

For En = Em we normalize them:

This we combine and write

(IEm), <1» = L bmiEn I<1» = (Em I<1». (4.5)

This is the expected identity (4.5d).

One can multiply both sides of (4.4d) with the operator H,

H<I> = L HIEn)(Enl<l» = L En IEn)(En I <1»,

(t/J, <1» = L (t/J IEn)(En I<1» = L (En It/J)*(Enl <1». (4.11d)

In particular if one chooses t/J = <I> one obtains

Equation (4.11d) is the analogue of the formula

for the ordinary scalar product in (R3.

As in the three-dimensional space 1R 3 , a vector ¢ is completely specified

LI¢il l < 00, (4.12d')

(HP¢, HP¢) = L (¢IHPIEn)(EnIHPI¢)

= LE;PI(Enl¢)ll<oo forany p=0,1,2, .... (4.1 3d)

(Ix), ¢) == fdY(,x), ly»<yl¢)·

This we rewrite with the following definition of the new symbol

(Ix>, ¢) = f dy <xly><yl¢>· (4.6c)

<xl¢> = Joo dy <xly><yl¢>. (4.6c')

The scalar products <y I¢ >are functions of the continuous variable y in

(Ernl¢) = L bEmdEnl¢) (4.6d')

where the Ix}!, are again eigenvectors of Q:

p{xlfjJ) = f p(y) dy p{xIY}pp{ylfjJ),

one has to demand

p{xIY}p = e~~») -ll5(X - y) = p-l(y)t5(X - y).

1= fdX,X)<X' = fp(X)dX'X}pp{X I. (4.9c)

which we call the spectral resolution of the self-adjoint operator Q with

("', fjJ) = f dx <"'lx)<xlfjJ)· (4. 11 c)

<xl¢) = ¢(x), <xlt/J)* = t/J*(x),

one can write (4.11c) in the form

(t/J, ¢) = f dx t/J*(x)¢(x). (4.llc')

In particular, if one chooses t/J = ¢ one obtains

An algebra.s;1 is called a -algebra if we have on the algebra a -operation

(ljJ, ¢) = L+oooo dx ljJ(x)¢(x) = L dz ljJ(z)¢(z)

f dx 1jJ(x)4>(x) = f dp \iI(p)4J(p), (S.40e)